From 208037a8138b37d618d6a8872e9286c4c5877875 Mon Sep 17 00:00:00 2001 From: "melodic-automation-lanes[bot]" <337595936+melodic-automation-lanes[bot]@users.noreply.github.com> Date: Mon, 26 Oct 2026 09:20:52 +0000 Subject: [PATCH 1/5] Resolve issue #8650 Lane-Activity: work-resolve-issue/resolve-issue --- ...eat-8650-ai-disclaimer-setting-5c2e9a41.md | 20 ++++++++++ docs/conventions/config-cascade/CHANGELOG.md | 3 ++ docs/conventions/config-cascade/README.md | 4 +- docs/conventions/loop-lane/CHANGELOG.md | 8 ++++ docs/conventions/loop-lane/README.md | 7 ++-- docs/conventions/work-items.yaml | 5 +++ plugins/work-items/.claude-plugin/plugin.json | 6 +++ plugins/work-items/README.md | 8 +++- plugins/work-items/reference/ai-disclaimer.md | 39 ++++++++++++++++--- .../work-items/reference/dogfood-filing.md | 6 +-- .../work-items/schemas/work-items.schema.json | 7 +++- .../work-items/skills/attend-queue/SKILL.md | 7 +++- .../skills/attend-queue/evals/evals.json | 18 ++++++++- plugins/work-items/skills/scan-todos/SKILL.md | 2 +- plugins/work-items/skills/triage/SKILL.md | 13 ++++--- .../skills/triage/context/apply-outcome.md | 4 +- .../work-items/skills/triage/evals/evals.json | 18 ++++++++- plugins/work-items/skills/work-loop/SKILL.md | 4 +- .../skills/work-loop/evals/evals.json | 29 +++++++++++++- .../work-loop/reference/objection-window.md | 10 ++++- plugins/work-items/skills/work/SKILL.md | 2 +- .../loop-lane-profile-claude-code-plugins.md | 5 ++- prompts/loops/loop-lane-prompts.md | 5 ++- scripts/affected-tests-no-suite.txt | 1 + 24 files changed, 190 insertions(+), 41 deletions(-) create mode 100644 .changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md create mode 100644 docs/conventions/work-items.yaml diff --git a/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md b/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md new file mode 100644 index 00000000000..2069128d30d --- /dev/null +++ b/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md @@ -0,0 +1,20 @@ +--- +bump: minor +--- + +### Added + +- **`ai_disclaimer` setting.** Set it per user in `userConfig` or per repository in + `docs/conventions/work-items.yaml`, which wins; the schema gains the key. + `reference/ai-disclaimer.md` states the key, its layers and the prefix once, and every reader + points there. + +### Changed + +- **Lanes no longer prefix the AI disclaimer by default.** Triage, the work-loop intake sweep, + attend-queue answers and autonomous filing from `work` and `scan-todos` add it only when + `ai_disclaimer` resolves `true`; set it to keep the old behavior. +- **The work-loop objection window no longer treats an escalation's answer as a new objection.** A + comment posted after a work-loop escalation marker and before the autonomous-eligible role label + was last applied answers that escalation, so an attend-queue answer without the disclaimer does + not re-escalate the item. diff --git a/docs/conventions/config-cascade/CHANGELOG.md b/docs/conventions/config-cascade/CHANGELOG.md index 8c1dddfc929..3b96e2ba11e 100644 --- a/docs/conventions/config-cascade/CHANGELOG.md +++ b/docs/conventions/config-cascade/CHANGELOG.md @@ -11,6 +11,9 @@ adding an optional layer or relaxing a rule additively is a minor bump. No contract change. +- **`work-items` gains `ai_disclaimer`.** The Implementers row, renamed `work-items` (settings + keys), adds the key beside the triage keys; this repository's `docs/conventions/work-items.yaml` + now exists and sets it `false`. - **`corrections` (`findings-buffer`) gains a user-scope adapter binding.** Its Implementers row declares two `userConfig` options, outside the file schema and read from user and managed settings only, that bind a repository's own adapter in that repository alone; the team file and diff --git a/docs/conventions/config-cascade/README.md b/docs/conventions/config-cascade/README.md index c025c61bec3..0873cc1657a 100644 --- a/docs/conventions/config-cascade/README.md +++ b/docs/conventions/config-cascade/README.md @@ -574,7 +574,7 @@ conformance. | `repo-fleet-hygiene` | `--config` then team then user-global | whole-file, no per-key | | `work-items` | overlay on the allowlist | per-key overlay | | `work-items` (recurring schedule) | team only | single-layer | -| `work-items` (triage keys) | team | per-key | +| `work-items` (settings keys) | team | per-key | | `songwriting` | project override over the bundled default | whole-file per template, no per-key | | `ai-slop` | later layer | per-key (lists replace) | | `docs-naming` | later layer; team on named policy-floor keys | per-key | @@ -647,7 +647,7 @@ its conformance cell. | `repo-fleet-hygiene` | no-floor | `.claude/repo-fleet-hygiene.conf` | user-global + team | `--config` then team then user-global | whole-file, no per-key | declared deviation; whole-file precedence (explicit `--config` > team > user-global fallback), no per-key merge, no overlay layer (#1099) | | `work-items` | no-floor | `.work-item-tracker.json` (repo root) | team + local overlay | overlay on the allowlist | per-key overlay | declared deviation ([ADR 0015](../../adr/0015-bind-the-tracker-at-repo-root-with-an-allowlisted-personal-overlay.md)): layers live at the repo root, not under `.claude/`; overlay (`.work-item-tracker.local.json`) merges per-key over a deny-by-default allowlist (lease TTL, jira/linear/gitea auth identity, `docs`); deliberately no user-global layer, since a cross-repo personal rung would reopen the per-user provider trap the allowlist forecloses. Anchors at the repo root (`CLAUDE_PROJECT_DIR`, else git toplevel), no CWD climb. The overlay's gitignore line is outside the `.claude/**/*.local.*` one-liner, so `/work-items:setup apply` appends it, announced, a declared exception to the recommend default (ADR 0015) | | `work-items` (recurring schedule) | no-floor | `.github/recurring-schedule.json` | team only | team only | single-layer | declared deviation ([ADR 0042](../../adr/0042-ratify-consumer-config-location-outliers-in-place.md)): the schedule lives under `.github/`, not `.claude/`, tracked and team-shared with no user-global layer and no overlay. Written by `track add` and `track recheck`; read by `track audit`, `done`, `due`, `search` and `stats`, by `work` candidate discovery, and by `setup`, which also writes it; anchors at `CLAUDE_PROJECT_DIR`, else git toplevel. Separate from the tracker binding row above | -| `work-items` (triage keys) | no-floor | team: `docs/conventions/work-items.yaml` (schema `plugins/work-items/schemas/work-items.schema.json`); per user: the plugin's `userConfig` option of the same name | `userConfig` + team | team | per-key | conforms to the [location axis](#location-of-the-team-layer-a-docs-convention-file-or-claudename) ([ADR 0064](../../adr/0064-home-plugin-customization-in-docs-conventions-yaml.md)), with the layer order of its Decision 8: the manifest default, then `userConfig`, then the team file; no `~/.claude` file and no overlay. Keys are `triage_repro_count` and `triage_objection_window_hours`. A value outside a key's range is named with its file, key and value, and that layer is dropped; the key then takes its manifest default, never a lower layer. The file is absent in this repository, so this row records a layer that is not tracked here. Read by `/work-items:triage` only, through `skills/triage/scripts/parse-concern-value.sh` from the repository root, or at a pinned ref under `--config-ref`; the `/work-items:work-loop` admission gate reads the window's end from the marker comment triage posts, not the file. Hand-edited, since `/work-items:setup apply` does not write this file yet | +| `work-items` (settings keys) | no-floor | team: `docs/conventions/work-items.yaml` (schema `plugins/work-items/schemas/work-items.schema.json`); per user: the plugin's `userConfig` option of the same name | `userConfig` + team | team | per-key | conforms to the [location axis](#location-of-the-team-layer-a-docs-convention-file-or-claudename) ([ADR 0064](../../adr/0064-home-plugin-customization-in-docs-conventions-yaml.md)), with the layer order of its Decision 8: the manifest default, then `userConfig`, then the team file; no `~/.claude` file and no overlay. Keys are `triage_repro_count`, `triage_objection_window_hours` and `ai_disclaimer`. A value outside a key's range is named with its file, key and value, and that layer is dropped; the key then takes its manifest default, never a lower layer. This repository's file sets `ai_disclaimer: false`. The triage keys are read by `/work-items:triage` only, through `skills/triage/scripts/parse-concern-value.sh` from the repository root, or at a pinned ref under `--config-ref`; the `/work-items:work-loop` admission gate reads the window's end from the marker comment triage posts, not the file. `ai_disclaimer` is read through the same script by each lane that follows `plugins/work-items/reference/ai-disclaimer.md`. Hand-edited, since `/work-items:setup apply` does not write this file yet | | `songwriting` | no-floor | `songwriting/templates/pat-pattison/.md` | team only, over a bundled default | project override over the bundled default | whole-file per template, no per-key | declared deviation ([ADR 0042](../../adr/0042-ratify-consumer-config-location-outliers-in-place.md)): the override path is under `songwriting/`, not `.claude/`. One tracked team file per template wins whole over `${CLAUDE_PLUGIN_ROOT}/context/pat-pattison/templates/.md`, the first match, and freezes that template against plugin updates; no user-global layer and no overlay. The ADR rules on location only. Written by `/songwriting:setup apply`, read by the craft skills | | `ai-slop` | no-floor | `.claude/ai-slop.json` | all three | later layer | per-key (lists replace) | conforms; per-key override, resolved by `/ai-slop:audit` (user-global, team, `.claude/ai-slop.local.json` overlay). Four list keys are additive-by-replacement rather than merged (`vocab_add` / `vocab_remove` tune the shipped word list, `phrase_add` / `phrase_remove` the shipped model-era phrase roster; the later layer's list wins per key). No policy-floor class: every key is a taste dial over prose style, and a personal overlay that silences a rule weakens nothing another surface depends on. Keys owned by `/ai-slop:setup`; `_comment` is an allowed free-text annotation, not drift | | `docs-naming` | team-only | `.claude/docs-naming.json` | all three | later layer; team on named policy-floor keys | per-key | conforms, with the one sanctioned dual-read declared by retirement records `docs-naming-r001` and `docs-naming-r002` (the retired `.claude/docs-hygiene.json` and `.claude/docs-hygiene.local.json` are read as authority, with a WARN, only while the new name is absent); per-key override on `file_names.*`, with a policy-floor class on `tiers`, `generated`, `sweep_exclude`, `sweep_exclude_sites`, and the three `exempt_*` keys: a personal layer may ADD entries and never remove them, and `generated` is team-layer only. Those keys decide what `/docs-naming:realign-file-names` does to a tree (which files are frozen, which reference forms are rewritten, and which shell command runs after a move), so narrowing one from a single machine would weaken a team decision, while adding a scope root or an exemption weakens nothing and stays open. `rule`, `regex`, and `redirect_map` are nearest-wins. Keys owned by [`plugins/docs-naming/reference/config.md`](../../../plugins/docs-naming/reference/config.md), which also partitions them from plugin `userConfig` (this plugin declares none, so a layer naming one is an inert unknown key). Written by `/docs-naming:setup apply`, resolved by `plugins/docs-naming/scripts/resolve-config.sh` | diff --git a/docs/conventions/loop-lane/CHANGELOG.md b/docs/conventions/loop-lane/CHANGELOG.md index c518b2e42b4..bea22222751 100644 --- a/docs/conventions/loop-lane/CHANGELOG.md +++ b/docs/conventions/loop-lane/CHANGELOG.md @@ -5,6 +5,14 @@ topology, the escalation contract, the capability-tier vocabulary, or any loop-l major bump, and additive guidance is a minor bump. A new model release re-audits the capability-tier table (§3); drift found by that audit is recorded here. +## [12.7.0] - 2026-10-26 + +Minor. Guidance only; no topology, escalation-contract, capability-tier or §4 invariant changed. + +- **AI disclaimer is a setting.** The attend-queue park and hold comments open with the AI + disclaimer only when work-items' `ai_disclaimer` setting resolves `true`; it defaults to `false`. + §4's lane-editable config adds that both lanes read `work-items.yaml` for the key (#8650). + ## [12.6.0] - 2026-10-26 Minor. Guidance only; no topology, escalation-contract, capability-tier or §4 invariant changed. diff --git a/docs/conventions/loop-lane/README.md b/docs/conventions/loop-lane/README.md index 83a2f6d59d5..104bf8a7175 100644 --- a/docs/conventions/loop-lane/README.md +++ b/docs/conventions/loop-lane/README.md @@ -900,7 +900,8 @@ offer to persist the answer; the headless path never waits on one. time that the lane's own process could edit (audited in #7680). work-loop and attend-queue read the tracker binding `.work-item-tracker.json` from the checkout, including the role labels that admission matches on. The triage step they run reads `docs/conventions/work-items.yaml` and -`docs/conventions/pr-pipeline.yaml` from the working tree when no `--config-ref` is given. The +`docs/conventions/pr-pipeline.yaml` from the working tree when no `--config-ref` is given, and +both lanes read that `work-items.yaml` for `ai_disclaimer` the same way. The babysit loop reads its `babysit_loop_*` keys from the `source-control` config seam's user-global, team and local layers, and reads the team layer from the working tree when the checkout is the target repository. A deny rule makes such a file read-only to a lane only when at least these four @@ -1210,8 +1211,8 @@ gh issue comment -R "$R" --body "Parked by operator: ." When an `attend-queue` session writes this comment, or the hold comment below, on the operator's behalf, it opens with the -[AI disclaimer](../../../plugins/work-items/reference/ai-disclaimer.md); the operator's own words -need none. +[AI disclaimer](../../../plugins/work-items/reference/ai-disclaimer.md) when that file's setting +resolves `true`; the operator's own words need none. `list-frontier --autonomous` excludes that label, so `work-loop` stops selecting the item from its next cycle-start snapshot. The comment carries no machine marker, which is what marks the item diff --git a/docs/conventions/work-items.yaml b/docs/conventions/work-items.yaml new file mode 100644 index 00000000000..c898f2c8760 --- /dev/null +++ b/docs/conventions/work-items.yaml @@ -0,0 +1,5 @@ +# yaml-language-server: $schema=../../plugins/work-items/schemas/work-items.schema.json +# Team layer for the work-items plugin's settings. +# Resolution: plugins/work-items/reference/ai-disclaimer.md (ai_disclaimer). +# Autonomous lane posts here come from a bot account, whose author line already marks them. +ai_disclaimer: false diff --git a/plugins/work-items/.claude-plugin/plugin.json b/plugins/work-items/.claude-plugin/plugin.json index 8faf32bece4..1ab9c056476 100644 --- a/plugins/work-items/.claude-plugin/plugin.json +++ b/plugins/work-items/.claude-plugin/plugin.json @@ -37,6 +37,12 @@ "default": 0, "min": 0 }, + "ai_disclaimer": { + "type": "boolean", + "title": "AI disclaimer on lane posts", + "description": "When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false. The repository's docs/conventions/work-items.yaml overrides this per repo.", + "default": false + }, "work_dispatch_concurrency_cap": { "type": "number", "title": "Autonomous dispatch concurrency cap", diff --git a/plugins/work-items/README.md b/plugins/work-items/README.md index fe833b5a5ae..b40d9455c0b 100644 --- a/plugins/work-items/README.md +++ b/plugins/work-items/README.md @@ -174,7 +174,7 @@ off) have a repository layer as well: the same keys in the consuming repo's `docs/conventions/work-items.yaml`, checked by [`schemas/work-items.schema.json`](schemas/work-items.schema.json), win over the per-user values. Lowest first: the manifest default, `userConfig`, the -repository file. Only `/work-items:triage` reads the file, through the bundled +repository file. Only `/work-items:triage` reads these two keys, through the bundled `parse-concern-value.sh`, and reports which layer supplied each value; the work-loop admission gate reads the window's end from the marker comment triage posts. A lane that passes `--config-ref ` to triage has the @@ -185,6 +185,11 @@ triage posts an "Objection window until " comment beside the brief of each item it makes autonomous-eligible; the work-loop lane skips the item until the window ends and escalates it if a person replies inside the window. +`ai_disclaimer` (default `false`) takes the same layers, in the same order: +when it resolves `true`, autonomous lanes start each tracker item and comment +they create with the AI disclaimer. The key, its layers and the prefix are +stated once in [`reference/ai-disclaimer.md`](reference/ai-disclaimer.md). + Everything else is project-specific behavior that routes through the consuming repo's own surfaces: the bound provider in `.work-item-tracker.json` (including the optional `config.role_labels` canonical-role → label remap), its labels @@ -226,6 +231,7 @@ reads it from. | `decompose_container_publish` | boolean | *(none)* | `CLAUDE_PLUGIN_OPTION_DECOMPOSE_CONTAINER_PUBLISH` | When true, /work-items:decompose pre-selects the spec-container offer in its approval round for multi-session breakdowns; the approval gate stays mandatory. Unset or false: the plain ask, with a default answer of no. No default is declared, so unset stays distinct from false. | | `triage_repro_count` | number
*min 1* | `2` | `CLAUDE_PLUGIN_OPTION_TRIAGE_REPRO_COUNT` | Independent reproductions /work-items:triage needs, each with evidence in its comment, before it marks a bug confirmed. Default 2. The repository's docs/conventions/work-items.yaml overrides this per repo. | | `triage_objection_window_hours` | number
*min 0* | `0` | `CLAUDE_PLUGIN_OPTION_TRIAGE_OBJECTION_WINDOW_HOURS` | Hours after triage makes an item autonomous-eligible during which the work-loop lane skips it, so a person can object in a reply. 0 (the default) is off. The repository's docs/conventions/work-items.yaml overrides this per repo. | +| `ai_disclaimer` | boolean | `false` | `CLAUDE_PLUGIN_OPTION_AI_DISCLAIMER` | When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false. The repository's docs/conventions/work-items.yaml overrides this per repo. | | `work_dispatch_concurrency_cap` | number
*min 1* | *(none)* | `CLAUDE_PLUGIN_OPTION_WORK_DISPATCH_CONCURRENCY_CAP` | Maximum worker rows /work-items:work's autonomous execute step lets /implementation:implement-dispatch run at once in one plan phase, passed as its --wave-cap ceiling. A fractional value is floored. Unset: implement-dispatch's implement_dispatch_wave_cap option, else its internal default. | | `lane_instance` | string | *(none)* | `CLAUDE_PLUGIN_OPTION_LANE_INSTANCE` | Writer identity for this machine's work-loop telemetry, the suffix of its telemetry marker, so concurrent lane instances never overwrite each other. Absent: the sanitized lowercased hostname. Must match ^\[a-z0-9\]\[a-z0-9-\]{0,31}$. It appears verbatim in tracker comments. | | `lane_app_slug` | string | *(none)* | `CLAUDE_PLUGIN_OPTION_LANE_APP_SLUG` | Slug of the GitHub App whose installation token the work-loop lane runs under, passed as --app-slug to its telemetry upsert so the write identity resolves to \[bot\]. Needed only under an App token. Absent: the identity resolves from gh api user. Must match ^\[a-z0-9\]\[a-z0-9-\]*$. | diff --git a/plugins/work-items/reference/ai-disclaimer.md b/plugins/work-items/reference/ai-disclaimer.md index c5ebf396e1c..d3bedc9216f 100644 --- a/plugins/work-items/reference/ai-disclaimer.md +++ b/plugins/work-items/reference/ai-disclaimer.md @@ -1,8 +1,29 @@ # AI disclaimer: the lane-neutral provenance prefix -When an autonomous lane creates a tracker item or comment without a human turn, prefix the body with -this disclaimer so readers know the text is model-generated. The `{lane}` token is the invoking lane's -short name (for example `triage`, `work-loop`, `scan-todos`, `babysit-prs`). +When an autonomous lane creates a tracker item or comment without a human turn, or an agent writes +one on an operator's behalf, and the `ai_disclaimer` setting resolves `true`, prefix the body with +this disclaimer so readers know the text is model-generated. When it resolves `false`, add no +prefix. The `{lane}` token is the invoking lane's short name (for example `triage`, `work-loop`, +`scan-todos`, `babysit-prs`). + +## Setting + +`ai_disclaimer` is a boolean, default `false`. Resolve it once per run, before the first post. Layers, +lowest first: the manifest default, then the per-user `userConfig` value, then the repository's +`docs/conventions/work-items.yaml` (schema: [`../schemas/work-items.schema.json`](../schemas/work-items.schema.json)), +the same order as `/work-items:triage`'s "Triage settings": + +1. **Repository.** From the repository root, run + `bash "/skills/triage/scripts/parse-concern-value.sh" docs/conventions/work-items.yaml ai_disclaimer`, + `` being this plugin's root directory, with `--ref ''` before the file argument + when the invocation carried a `--config-ref` that triage's check accepted. Non-empty output is + the value, from the repository layer. +2. **Per user.** When the output is empty (no file or no key), the value is the + `${user_config.ai_disclaimer}` the invoking skill's body renders (`triage`, `attend-queue`, + `work-loop`, `work` and `scan-todos` render it). A surviving literal `${user_config.…}` + placeholder, or a skill that renders none, means unset: take the default. +3. **Check the value.** Anything other than `true` or `false` names its source and the value, and + drops that layer to the default, never to a lower layer. It never stops the run. ## Canonical form @@ -15,9 +36,15 @@ when the provenance phase differs. Triage's `during triage` form is correct only ## Where this applies -Every cross-lane self-observation and autonomous mutation contract points here rather than restating -the sentence: +Every cross-lane self-observation and autonomous mutation contract points here for both the +condition and the sentence, rather than restating either: -- [`dogfood-filing.md`](dogfood-filing.md): autonomous filing sequence +- [`dogfood-filing.md`](dogfood-filing.md): autonomous filing sequence, from `work` and `scan-todos` - [`../skills/triage/SKILL.md`](../skills/triage/SKILL.md): triage mutations +- [`../skills/attend-queue/SKILL.md`](../skills/attend-queue/SKILL.md): answers written on the operator's behalf +- [`../skills/work-loop/SKILL.md`](../skills/work-loop/SKILL.md): the intake sweep - Loop-lane standing prompts and lane-absorption skills: reconcile against this contract + +A reader that tells lane text from a person's by the prefix (the work-loop objection window) cannot +rely on it while the setting resolves `false`, so it does not rest on the prefix alone: the +objection window also reads the author and when the item was returned to the frontier. diff --git a/plugins/work-items/reference/dogfood-filing.md b/plugins/work-items/reference/dogfood-filing.md index d1c0d28dc18..f7ac2fb6000 100644 --- a/plugins/work-items/reference/dogfood-filing.md +++ b/plugins/work-items/reference/dogfood-filing.md @@ -72,9 +72,9 @@ creating it ([`../skills/track/actions/add.md`](../skills/track/actions/add.md) gate"). On an **autonomous lane**, a `/loop` or `/schedule` session whose standing rules already authorize tracker mutations, those standing rules **are** the authorization, the same resolution triage's direction gate makes ([`../skills/triage/SKILL.md`](../skills/triage/SKILL.md) "Direction -gate"). An autonomous lane prefixes every item and comment it creates with the lane-neutral AI -disclaimer ([`ai-disclaimer.md`](ai-disclaimer.md)), substituting this lane's short name for -`{lane}`. +gate"). Whether an autonomous lane prefixes each item and comment it creates with the lane-neutral +AI disclaimer, and its form, are in [`ai-disclaimer.md`](ai-disclaimer.md); substitute this lane's +short name for `{lane}`. ## Reconciliation note diff --git a/plugins/work-items/schemas/work-items.schema.json b/plugins/work-items/schemas/work-items.schema.json index 7f1678b9f25..55a9b6922d3 100644 --- a/plugins/work-items/schemas/work-items.schema.json +++ b/plugins/work-items/schemas/work-items.schema.json @@ -2,7 +2,7 @@ "$schema": "https://json-schema.org/draft/2020-12/schema", "$id": "https://raw.githubusercontent.com/melodic-software/claude-code-plugins/main/plugins/work-items/schemas/work-items.schema.json", "title": "work-items team config", - "description": "Schema for a repository's docs/conventions/work-items.yaml, the team layer of the work-items plugin's settings (ADR 0064). A key set here wins over the same key in the plugin's userConfig. Reader: /work-items:triage only; the /work-items:work-loop admission gate reads the objection window's end from the marker comment triage posts.", + "description": "Schema for a repository's docs/conventions/work-items.yaml, the team layer of the work-items plugin's settings (ADR 0064). A key set here wins over the same key in the plugin's userConfig. Readers: /work-items:triage for the triage keys (the /work-items:work-loop admission gate reads the objection window's end from the marker comment triage posts), and every lane that follows reference/ai-disclaimer.md for ai_disclaimer.", "type": "object", "additionalProperties": false, "properties": { @@ -18,6 +18,11 @@ "minimum": 0, "default": 0, "description": "Hours after triage marks an item autonomous-eligible during which the work-loop lane does not dispatch it, so a person can object. 0 turns the window off." + }, + "ai_disclaimer": { + "type": "boolean", + "default": false, + "description": "Whether autonomous work-items lanes start each tracker item and comment they create with the AI disclaimer (reference/ai-disclaimer.md). false adds no prefix." } } } diff --git a/plugins/work-items/skills/attend-queue/SKILL.md b/plugins/work-items/skills/attend-queue/SKILL.md index 0dd74c9b5b4..5f582fbc153 100644 --- a/plugins/work-items/skills/attend-queue/SKILL.md +++ b/plugins/work-items/skills/attend-queue/SKILL.md @@ -243,8 +243,11 @@ answer without stopping the pass to ask which item is in front of them. completion route for this lane to hand it to, and the human-gated role label is what keeps it in front of an operator. -Answers and dispositions written by the agent on the operator's behalf carry triage's AI -disclaimer; the operator's own words need none. +Answers and dispositions written by the agent on the operator's behalf carry the AI disclaimer +when the setting resolves `true` +([`${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md`](${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md), +`{lane}` → `attend-queue`; the per-user value it reads is `${user_config.ai_disclaimer}`); the +operator's own words need none. The operator reads those answers between rows, so each one opens with the decision and stops there: invoke `/writing:be-concise` via the Skill tool when the `writing` plugin is installed; diff --git a/plugins/work-items/skills/attend-queue/evals/evals.json b/plugins/work-items/skills/attend-queue/evals/evals.json index 697b5b96877..4c90855e0b9 100644 --- a/plugins/work-items/skills/attend-queue/evals/evals.json +++ b/plugins/work-items/skills/attend-queue/evals/evals.json @@ -19,9 +19,10 @@ { "id": 2, "name": "attend-queue-answers-write-back-ratify-and-flip", - "prompt": "/work-items:attend-queue\n\nThe operator works three rows: an [escalated] item where the worker asked which of two API shapes to keep; a [ratify] row for a C3 bug-fix admission; and an [intake] bug report. The operator answers the API question, ratifies the C3 admission, and directs triage on the bug.", - "expected_output": "For each row, the session acquires the seam claim (assignee + lease) before any comment, label, or triage mutation. For the escalated row, the decision is driven via /planning:interview when the planning plugin is installed (inline focused questions, one at a time, as the documented fallback) and the answer is written back as an issue comment on the item — the decision lives on the tracker, not only in the session — after which the item is flipped in one edit while the claim is still held: the autonomous-eligible role label applied and the human-gated role label removed, both resolved from config.role_labels, then @me cleared from assignees via the adapter assignee edit. For the ratify row, the ratification is recorded as a reply comment and the same flip-while-claimed then assignee-clear sequence applies; the item re-enters the worker loop's frontier — this lane never dispatches or executes it. For the intake row, /work-items:triage runs under its interactive direction gate: recommendation first, mutation only after the operator's direction. Agent-authored comments carry the AI disclaimer.", + "prompt": "/work-items:attend-queue\n\nThe operator works three rows: an [escalated] item where the worker asked which of two API shapes to keep; a [ratify] row for a C3 bug-fix admission; and an [intake] bug report. The operator answers the API question, ratifies the C3 admission, and directs triage on the bug. The per-user ai_disclaimer is true and the repository has no docs/conventions/work-items.yaml.", + "expected_output": "For each row, the session acquires the seam claim (assignee + lease) before any comment, label, or triage mutation. For the escalated row, the decision is driven via /planning:interview when the planning plugin is installed (inline focused questions, one at a time, as the documented fallback) and the answer is written back as an issue comment on the item — the decision lives on the tracker, not only in the session — after which the item is flipped in one edit while the claim is still held: the autonomous-eligible role label applied and the human-gated role label removed, both resolved from config.role_labels, then @me cleared from assignees via the adapter assignee edit. For the ratify row, the ratification is recorded as a reply comment and the same flip-while-claimed then assignee-clear sequence applies; the item re-enters the worker loop's frontier — this lane never dispatches or executes it. For the intake row, /work-items:triage runs under its interactive direction gate: recommendation first, mutation only after the operator's direction. Since ai_disclaimer resolves true from the per-user layer, agent-authored comments carry the AI disclaimer.", "files": [], + "narration": true, "expectations": [ "Acquires the seam claim before mutating each row — assignee + lease, same protocol as /work-items:work", "Uses /planning:interview behind an installed-ness gate with the inline focused-questions fallback documented at the site", @@ -158,6 +159,19 @@ "Applies neither the autonomous-eligible role label nor a work-class label to #980", "Does not merge #980" ] + }, + { + "id": 12, + "name": "attend-queue-writes-answer-without-disclaimer-when-setting-off", + "prompt": "/work-items:attend-queue\n\nThe repository's docs/conventions/work-items.yaml sets ai_disclaimer: false. The operator works one [escalated] row, #612 \"Exporter drops the trailing newline\", which carries work-class: mechanical and whose marker asks whether to keep or strip the trailing newline; the operator answers: keep it.", + "expected_output": "After the brief-before-ask restatement and the operator's answer, the session holds the seam claim, writes the answer back as a comment on #612 and flips the item. ai_disclaimer resolves false from the repository layer, so the answer comment it writes on the operator's behalf carries no AI disclaimer prefix; it opens with the decision.", + "files": [], + "narration": true, + "expectations": [ + "Resolves ai_disclaimer to false from the repository's docs/conventions/work-items.yaml", + "Writes the operator's answer back as a comment on #612 with no AI disclaimer prefix", + "Opens the answer comment with the decision (keep the trailing newline)" + ] } ] } diff --git a/plugins/work-items/skills/scan-todos/SKILL.md b/plugins/work-items/skills/scan-todos/SKILL.md index e914f68d22d..e2056492c49 100644 --- a/plugins/work-items/skills/scan-todos/SKILL.md +++ b/plugins/work-items/skills/scan-todos/SKILL.md @@ -65,7 +65,7 @@ git -C "${CLAUDE_PROJECT_DIR:-$(git rev-parse --show-toplevel)}" grep -nE '\b(TO 1. **For each marker**, read context (10 lines before/after), then classify: - **Resolve now**, small fix; do the work, remove the marker -- **File a work item + remove marker**, significant work; create it by invoking `/work-items:track add` via the Skill tool following the shared self-observation contract ([`${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md`](${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md): dedupe → categorize → fixed shape → `needs-triage`), remove the inline marker (do not leave `TODO` as a stand-in for the item) +- **File a work item + remove marker**, significant work; create it by invoking `/work-items:track add` via the Skill tool following the shared self-observation contract ([`${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md`](${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md): dedupe → categorize → fixed shape → `needs-triage`; per-user `ai_disclaimer`: `${user_config.ai_disclaimer}`), remove the inline marker (do not leave `TODO` as a stand-in for the item) - **Remove (already done)**, work completed; delete the comment - **False positive**, structured grammar or external upstream citation misclassified; fix the exclusion if systemic, otherwise note it in the PR diff --git a/plugins/work-items/skills/triage/SKILL.md b/plugins/work-items/skills/triage/SKILL.md index 7518e0d50fd..9d086db1679 100644 --- a/plugins/work-items/skills/triage/SKILL.md +++ b/plugins/work-items/skills/triage/SKILL.md @@ -113,7 +113,7 @@ Report a `blocked by won't-do` item for the operator to decide (drop the edge, r Within each bucket, list the oldest item first, one line per item. -List open items and filter into buckets programmatically (adapter: "List items", bare read). Apply the lane-infrastructure exclusion ("Scope: raw intake only") to that listing **before** bucketing, so a telemetry issue carrying the raw marker is filtered out rather than bucketed under it. The `blocked by won't-do` bucket reads the count from the listing rows, whatever labels the item carries: an item already triaged still lands there, and one that also fits an earlier bucket is listed once, under the earlier bucket. **Defensive skip** (the first three buckets only): drop any item that already carries a native `blocked-by` edge *and* a prior triage comment (machine disclaimer or a needs-info comment from an earlier pass, in any shape it was posted in), a stray re-label from another lane must not cost a full re-investigation. When the repo accepts requests in the form of outside PRs, list them in the same buckets with a `[PR]` or `[issue]` prefix on every line. Only PRs from outside contributors appear: a PR a collaborator is still working on is their work, not intake. That limit applies to this listing only; a PR named explicitly gets triaged whoever opened it. Present as a compact table. +List open items and filter into buckets programmatically (adapter: "List items", bare read). Apply the lane-infrastructure exclusion ("Scope: raw intake only") to that listing **before** bucketing, so a telemetry issue carrying the raw marker is filtered out rather than bucketed under it. The `blocked by won't-do` bucket reads the count from the listing rows, whatever labels the item carries: an item already triaged still lands there, and one that also fits an earlier bucket is listed once, under the earlier bucket. **Defensive skip** (the first three buckets only): drop any item that already carries a native `blocked-by` edge *and* a prior triage comment (a verification, needs-info or other triage comment from an earlier pass, with or without the AI disclaimer, in any shape it was posted in), a stray re-label from another lane must not cost a full re-investigation. When the repo accepts requests in the form of outside PRs, list them in the same buckets with a `[PR]` or `[issue]` prefix on every line. Only PRs from outside contributors appear: a PR a collaborator is still working on is their work, not intake. That limit applies to this listing only; a PR named explicitly gets triaged whoever opened it. Present as a compact table. ### Board page @@ -199,9 +199,9 @@ Classify **bug vs enhancement** first. It steers the rest of the flow (bugs get **Direction gate.** Recommending is read-only; the gate governs *mutation*, labels, comments, closes, item creation, and which side of it you are on is fixed by how triage was invoked: - **Interactive session**, a human operator is present and no standing lane rules were supplied. **Brief before asking**: before presenting the recommendation, restate (1) which item (number + one-line title), (2) the decision being asked, and (3) the consequence of each option **you present**, the recommendation and the alternatives you are actually putting to the operator, not every target state the state machine admits, then present the recommendation and **wait for the user's explicit direction** before mutating anything. This is the default whenever the invocation carries no autonomous mandate. The restatement is not optional compression fodder: a terse output style must never drop it, and it applies on every decision question, not only the first one of a pass, the operator working several rows in sequence (e.g. via `/work-items:attend-queue`) cannot be assumed to still be holding a prior item's context. -- **Autonomous lane**. Triage is running unattended as a `/loop` or `/schedule` AFK session whose **lane standing directive**, the text supplied with its `/loop` / `/schedule` invocation that authorizes triage mutations. Already satisfies the direction gate. This is **not** the `discipline` plugin's sense of "standing rules" (project-configured rules in consumer settings); here the lane directive **is** the direction this gate requires: treat the gate as satisfied and proceed through verification and outcome without a human turn, prefixing every comment and item you create with the AI disclaimer. A general mandate such as "handle routine work" counts only when it explicitly authorizes triage label/comment mutations; otherwise fall back to the interactive branch. There is no operator turn to wait for, so blocking here would deadlock the lane, the gate is met by the lane's mandate, not skipped. +- **Autonomous lane**. Triage is running unattended as a `/loop` or `/schedule` AFK session whose **lane standing directive**, the text supplied with its `/loop` / `/schedule` invocation that authorizes triage mutations. Already satisfies the direction gate. This is **not** the `discipline` plugin's sense of "standing rules" (project-configured rules in consumer settings); here the lane directive **is** the direction this gate requires: treat the gate as satisfied and proceed through verification and outcome without a human turn, with each comment and item you create carrying the AI disclaimer when the setting in "AI disclaimer" below resolves `true`. A general mandate such as "handle routine work" counts only when it explicitly authorizes triage label/comment mutations; otherwise fall back to the interactive branch. There is no operator turn to wait for, so blocking here would deadlock the lane, the gate is met by the lane's mandate, not skipped. -The autonomous branch is the mode the AI disclaimer already anticipates: a session that mutates without a human turn. The two are one mode, not a contradiction. +The autonomous branch is the mode the AI disclaimer anticipates: a session that mutates without a human turn. The two are one mode, not a contradiction. A lane triaging many items holds to the provider's API budget ([`tracker-seam.md`](${CLAUDE_PLUGIN_ROOT}/reference/tracker-seam.md) "Gotchas"): its writes go one at a time, and a rate-limit error stops the run's tracker calls, never a retry. @@ -265,6 +265,7 @@ admission gate enforces the window. ## AI disclaimer -When creating comments or items during autonomous/agent triage sessions, prefix with the canonical -form in [`${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md`](${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md) -(`{lane}` → `triage`). +Whether comments and items created during autonomous/agent triage sessions carry the AI disclaimer, +and its canonical form, are in +[`${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md`](${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md) +(`{lane}` → `triage`; the per-user value it reads is `${user_config.ai_disclaimer}`). diff --git a/plugins/work-items/skills/triage/context/apply-outcome.md b/plugins/work-items/skills/triage/context/apply-outcome.md index 995d379a602..a2ae010d937 100644 --- a/plugins/work-items/skills/triage/context/apply-outcome.md +++ b/plugins/work-items/skills/triage/context/apply-outcome.md @@ -118,7 +118,7 @@ outcome or lane edit. Step 3 posts it on a passed verification, before the label edit that applies `status: confirmed`. If the comment cannot be posted, apply no label. In an autonomous session it starts with the AI -disclaimer. +disclaimer when the setting resolves `true` ([`reference/ai-disclaimer.md`](../../../reference/ai-disclaimer.md)). ```markdown **Verified: ** @@ -139,7 +139,7 @@ checked and what it showed. Posted with the brief by every outcome that applies the autonomous-eligible role label, when `triage_objection_window_hours` resolved above 0. `` is the post time plus that many hours, in UTC, written `YYYY-MM-DDTHH:MM:SSZ`. The marker line is first so the work-loop admission gate can -find it; in an autonomous session the AI disclaimer follows it. +find it; in an autonomous session the AI disclaimer, when the setting adds one, follows it. Post it before the label edit that applies the autonomous-eligible role label. If the comment cannot be posted, apply no label at all: the item keeps its raw marker and the next sweep retries diff --git a/plugins/work-items/skills/triage/evals/evals.json b/plugins/work-items/skills/triage/evals/evals.json index 4e9b7347ef0..9b0da3b7705 100644 --- a/plugins/work-items/skills/triage/evals/evals.json +++ b/plugins/work-items/skills/triage/evals/evals.json @@ -71,9 +71,10 @@ { "id": 6, "name": "triage-autonomous-lane-treats-standing-rules-as-direction", - "prompt": "/work-items:triage 71 — running unattended as a /loop AFK lane whose standing rules authorize triage label and comment mutations; #71 is an external bug report you have just reproduced", - "expected_output": "Recognizes the autonomous branch of the direction gate: the lane's standing rules ARE the direction the gate requires, so it proceeds through verification and outcome without waiting for a human turn rather than blocking (which would deadlock the lane). Applies the outcome's label and comment writes and prefixes every comment and item it creates with the AI disclaimer. Does not fall back to the interactive wait-gate, which applies only when a human operator is present and no standing lane rules were supplied.", + "prompt": "/work-items:triage 71 — running unattended as a /loop AFK lane whose standing rules authorize triage label and comment mutations; the repository's docs/conventions/work-items.yaml sets ai_disclaimer: true; #71 is an external bug report you have just reproduced", + "expected_output": "Recognizes the autonomous branch of the direction gate: the lane's standing rules ARE the direction the gate requires, so it proceeds through verification and outcome without waiting for a human turn rather than blocking (which would deadlock the lane). Applies the outcome's label and comment writes and, since ai_disclaimer resolves true from the repository layer, prefixes every comment and item it creates with the AI disclaimer. Does not fall back to the interactive wait-gate, which applies only when a human operator is present and no standing lane rules were supplied.", "files": [], + "narration": true, "expectations": [ "Selects the autonomous-lane branch of the direction gate — the /loop standing rules satisfy the direction requirement, so mutation proceeds without a human turn", "Does NOT wait for explicit user direction — that gate is the interactive-session default, not the autonomous-lane path; blocking here would deadlock the lane", @@ -448,6 +449,19 @@ "Applies needs-human and status:ready and clears the raw-intake marker", "Comments naming /work-items:decompose as the next step and does not run it" ] + }, + { + "id": 34, + "name": "triage-autonomous-lane-posts-without-disclaimer-when-setting-off", + "prompt": "/work-items:triage 72 — running unattended as a /loop AFK lane whose standing rules authorize triage label and comment mutations. The repository's docs/conventions/work-items.yaml sets ai_disclaimer: false, and your per-user ai_disclaimer is true. #72 is an external bug report you have just reproduced twice and are posting the verification comment for.", + "expected_output": "Resolves ai_disclaimer from its layers, the repository file winning over the per-user value, so it resolves false. Posts the verification comment and the outcome's comments without the AI disclaimer prefix, the verification comment starting with its **Verified: bug reproduced** line, and still proceeds through the outcome without waiting for a human turn.", + "files": [], + "narration": true, + "expectations": [ + "Resolves ai_disclaimer to false because the repository's docs/conventions/work-items.yaml wins over the per-user true", + "Posts the verification comment and every other comment it creates without the AI disclaimer prefix", + "Proceeds through the outcome without waiting for a human turn" + ] } ] } diff --git a/plugins/work-items/skills/work-loop/SKILL.md b/plugins/work-items/skills/work-loop/SKILL.md index 672a131eb45..47ed528fa50 100644 --- a/plugins/work-items/skills/work-loop/SKILL.md +++ b/plugins/work-items/skills/work-loop/SKILL.md @@ -227,7 +227,9 @@ the former default-on pause). **reported, never chased** (per the convention). 2. **Intake sweep.** Invoke `/work-items:triage` via the Skill tool over untriaged intake in its **autonomous lane**, this loop's launch-prompt standing rules are the direction its mutation gate requires, and every - comment or item it creates carries the AI disclaimer. Sweep hardening: an advisory issue + comment or item it creates carries the AI disclaimer when the setting resolves `true` + ([`${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md`](${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md); + the per-user value it reads is `${user_config.ai_disclaimer}`). Sweep hardening: an advisory issue authored by a workflow bot routes to the human-gated role label by default (this also lets drain exits terminate against automated intake). Applied together with a machine-marked `kind=routed-advisory` escalation comment (step 5's marker shape and record write), so the diff --git a/plugins/work-items/skills/work-loop/evals/evals.json b/plugins/work-items/skills/work-loop/evals/evals.json index 4803334fa26..9a069745ccf 100644 --- a/plugins/work-items/skills/work-loop/evals/evals.json +++ b/plugins/work-items/skills/work-loop/evals/evals.json @@ -4,8 +4,8 @@ { "id": 1, "name": "work-loop-admission-gate-fail-closed-with-hard-gates", - "prompt": "/work-items:work-loop\n\nRunning as a /loop autonomous lane whose standing rules authorize triage mutations. The frontier snapshot contains: a C2 mechanical doc fix; a C3 bug-fix-shaped item; a C3 feature-shaped item; an unclassified item; and a C2-looking dependency bump that recomputes a pinned checksum. A workflow bot filed two new advisory issues mid-cycle. The lane's telemetry issue does not carry the loop-lane: first-drain-ratified label.", - "expected_output": "The cycle re-reads durable loop state from the telemetry comment first, takes a cycle-start snapshot, and sweeps intake via /work-items:triage in its autonomous lane (standing rules satisfy the direction gate; AI disclaimer on comments). Admission is fail-closed: the C2 item is admitted; the C3 bug-fix item is NOT dispatched because first-drain-ratified.sh does not exit 0 (the label is absent) and it carries no recorded ratification \u2014 it is queued with the human-gated role label (resolved from config.role_labels) plus a machine-marked kind=ratify-c3 comment; a C3 item already carrying attend-queue's ratification reply would dispatch instead of being re-queued; the C3 feature-shaped and unclassified items are human-gated; the checksum-recomputation bump is human-gated by the path/topic hard gate regardless of its C2 surface reading. Bot-authored advisories route to the human-gated role by default and, arriving mid-cycle, are reported, never chased this cycle. Escalations use the role label plus the machine-marked escalation comment, never a new label. The loop authors PRs only and never merges.", + "prompt": "/work-items:work-loop\n\nRunning as a /loop autonomous lane whose standing rules authorize triage mutations. The frontier snapshot contains: a C2 mechanical doc fix; a C3 bug-fix-shaped item; a C3 feature-shaped item; an unclassified item; and a C2-looking dependency bump that recomputes a pinned checksum. A workflow bot filed two new advisory issues mid-cycle. The lane's telemetry issue does not carry the loop-lane: first-drain-ratified label. The repository's work-items settings file sets ai_disclaimer: true.", + "expected_output": "The cycle re-reads durable loop state from the telemetry comment first, takes a cycle-start snapshot, and sweeps intake via /work-items:triage in its autonomous lane (standing rules satisfy the direction gate; AI disclaimer on comments, since ai_disclaimer resolves true). Admission is fail-closed: the C2 item is admitted; the C3 bug-fix item is NOT dispatched because first-drain-ratified.sh does not exit 0 (the label is absent) and it carries no recorded ratification \u2014 it is queued with the human-gated role label (resolved from config.role_labels) plus a machine-marked kind=ratify-c3 comment; a C3 item already carrying attend-queue's ratification reply would dispatch instead of being re-queued; the C3 feature-shaped and unclassified items are human-gated; the checksum-recomputation bump is human-gated by the path/topic hard gate regardless of its C2 surface reading. Bot-authored advisories route to the human-gated role by default and, arriving mid-cycle, are reported, never chased this cycle. Escalations use the role label plus the machine-marked escalation comment, never a new label. The loop authors PRs only and never merges.", "files": [], "expectations": [ "Re-reads the machine-readable durable loop state from the telemetry comment at cycle start rather than trusting conversation context", @@ -463,6 +463,31 @@ "Changes no label on #43", "Continues the cycle rather than stopping" ] + }, + { + "id": 33, + "name": "work-loop-intake-sweep-posts-without-disclaimer-when-setting-off", + "prompt": "/work-items:work-loop\n\nRunning as a /loop autonomous lane whose standing rules authorize triage mutations. ai_disclaimer is unset per user and the repository has no docs/conventions/work-items.yaml. The cycle-start snapshot holds one untriaged item, #88, a clear documentation typo report.", + "expected_output": "The intake sweep runs /work-items:triage in its autonomous lane on #88. ai_disclaimer resolves false from the manifest default, so the comments and items the sweep creates carry no AI disclaimer prefix. The sweep's mutations otherwise proceed as usual.", + "files": [], + "narration": true, + "expectations": [ + "Runs the intake sweep via /work-items:triage in its autonomous lane on #88", + "Resolves ai_disclaimer to its default, false, since no layer sets it", + "Creates the sweep's comments and items with no AI disclaimer prefix" + ] + }, + { + "id": 34, + "name": "work-loop-escalation-answer-is-not-a-new-objection", + "prompt": "/work-items:work-loop\n\nA cycle starts at 2026-10-05T12:00:00Z. ai_disclaimer resolves false. #306 is agent-ready and work-class: mechanical with no open closing PR. Its objection window comment from the seam's write identity ended at 2026-10-04T10:00:00Z. At 2026-10-04T11:00:00Z the write identity posted a lane=work-loop kind=escalated marker for a maintainer's objection. At 2026-10-05T09:00:00Z the operator's account (write permission, not the write identity, no AI disclaimer) commented 'Decision: keep the old config format.' At 09:01 the same operator applied agent-ready; label-provenance reports that application as trusted. No comment follows it.", + "expected_output": "The operator's 09:00 comment sits after the work-loop escalation marker and before the latest trusted application of agent-ready, so it answers that escalation and is not a new objection; the maintainer's earlier objection is answered by the marker. With the window over, #306 continues to classification and is not escalated again.", + "files": [], + "expectations": [ + "Does not treat the operator's 09:00 comment as a new objection", + "Does not escalate #306 again", + "Continues #306 to classification" + ] } ] } diff --git a/plugins/work-items/skills/work-loop/reference/objection-window.md b/plugins/work-items/skills/work-loop/reference/objection-window.md index 130c5288e0d..7fefa7769be 100644 --- a/plugins/work-items/skills/work-loop/reference/objection-window.md +++ b/plugins/work-items/skills/work-loop/reference/objection-window.md @@ -17,9 +17,15 @@ a window. Check in this order: already answered by a later `lane=work-loop` escalation marker on the item. An escalation marker answers an objection only when the seam's configured write identity posted it, as with the window marker; the same marker text from any other author is an ordinary comment, so a - pasted marker cannot cancel an objection (and, posted by a person, is itself an objection). A person here is an + pasted marker cannot cancel an objection (and, posted by a person, is itself an objection). + A comment posted after such a marker and before the latest application of the + autonomous-eligible role label (its `at` in the labeler precondition's `label-provenance` read, + from a trusted labeler) answers that escalation, as `/work-items:attend-queue` writes its + answers, and is not a new objection. A person here is an author that is not the seam's write identity, is not a bot account (the tracker reports its - user type as `Bot`, or its login ends in `[bot]`), and whose comment carries no AI disclaimer. Step 5 escalates the item as + user type as `Bot`, or its login ends in `[bot]`), and whose comment carries no AI disclaimer + (a lane adds one only when the setting in + [`../../../reference/ai-disclaimer.md`](../../../reference/ai-disclaimer.md) resolves `true`). Step 5 escalates the item as `kind=escalated`, whose one-line question names the objecting comment, and it leaves the autonomous frontier. The cycle report lists it as `objection: # -> escalated`. 2. **Window still open** (`` after the cycle-start time): skip the candidate this cycle. It is diff --git a/plugins/work-items/skills/work/SKILL.md b/plugins/work-items/skills/work/SKILL.md index 76a6cfa5e33..e2b0309f4b3 100644 --- a/plugins/work-items/skills/work/SKILL.md +++ b/plugins/work-items/skills/work/SKILL.md @@ -312,7 +312,7 @@ On user confirmation ("yes"): 1. **Open the PR. Orchestrator-owned.** After the worker returns and the pre-PR diff gate passes, the **orchestrator** opens the PR. It is never the worker's to open (opening it from a worker would make the pre-PR gate a no-op, and `/implementation:implement-dispatch` already keeps PR creation out of every worker brief). The worker committed and pushed inside its own worktree, so the orchestrator, when `/source-control:pull-request` is among the available skills, invokes the **PR-only entry** `/source-control:pull-request create --pushed --worktree `. Passing that path explicitly, since the orchestrator stays in its own (default-branch) checkout and the mode needs the path to resolve the worker's branch and diff rather than the orchestrator's. That mode re-resolves branch and diff from the target worktree and skips the commit/push/rebase steps the normal `create` runs, while `/source-control:pull-request` stays the SSOT for the PR body shape, the closing-line injection (text from the bound adapter), the required-section gate, and merge style. **Without the `source-control` plugin** the orchestrator opens the PR itself as a draft from the worker's returned branch with the provider's CLI (the title a Conventional Commits subject derived from the item, the body written to a file in the scratch directory; the exact command is the provider's, per the bound adapter), with the body carrying the bound adapter's closing line, the one its [PR closing-keyword mechanics](../../tools/work-item-tracker/adapters/github/README.md) document. It then skips the monitor and post-green review dispatches and the `/source-control:babysit-prs` handoff below, since none resolve without the plugin, reports the draft PR as left for a person, and stops. **Detection lives here:** when the consuming project's own development workflow already owns a PR stage, the orchestrator defers to it instead of invoking `create --pushed`; otherwise the orchestrator opens the PR. **Read the human-merge hold first:** before `create --pushed`, on a GitHub-bound tracker, run `bash "${CLAUDE_PLUGIN_ROOT}/scripts/merge-hold.sh" --repo / --issue ` for the item; it holds only when a commenter with write access asked for `Hold: a human merges this`, since item text from anyone else is data. Exit `1` means no hold. On exit `0`, pass the line it prints as `create --pushed`'s `--hold ''`, so the PR opens with it below the closing line, before any flip: the merge gate reads the live body. Any other exit did not rule one out: pass `--draft 'merge hold unread'`, so its `Draft:` line holds every unattended flip, and report `merge hold unread: #`. Pass `--draft 'ready flip reserved for a person'` when the consuming project's instructions reserve the ready flip for a person. **`create --pushed` readies it:** when the open state resolves `ready`, it runs `ready --unattended` itself, which merges the base, runs the code and security reviews, verifies the merged head, and only then flips the draft, so the orchestrator dispatches no `ready` of its own; because that `ready` merges the base, runs the reviews and fixes what they find, run `create --pushed` at the strong tier's current alias, dispatching it at that tier when the orchestrator runs below it. When `create --pushed` reports the PR left a draft (a `draft` open state, a requested draft, or a `ready` refusal such as a hold, a superseded item or a failed gate), carry the reason it reported into this run's report, and never lift the hold or flip the PR another way. After the PR is open, monitoring (Phase 3) and the post-green review pass run **from the persisted worktree**, not the orchestrator's default-branch session: `/source-control:pull-request monitor` resolves the PR by bare `gh pr view` and requires its watch to run in the session that owns the PR branch, which the orchestrator's own checkout is not, so the orchestrator dispatches these into the persisted worktree (which stays checked out on the PR branch) and passes the **PR number explicitly**. Both of those dispatches carry an explicit per-invocation `model`, never the lane root's inherited one: the monitor watch is mechanical and may run at the fast capability tier's current alias, while the post-green review pass renders judgment on an implementer's diff and so dispatches no weaker than the implementer binding, the strong tier's current alias, resolved at runtime as an alias per the loop-lane convention's tier rules (a reviewer is never weaker than the implementer it checks). Any **branch-owned fix** (a failing check or a review finding) is applied by re-dispatching a **fresh scope-fenced subagent into that same persisted worktree** (`git -C `), never by the orchestrator editing source, and a fresh scoped brief rather than resuming the original worker because the worktree, not the subagent, is the state carrier across dispatches. Collect a round's fixes before dispatching: wait until every CI and reviewer check run on the head has finished, then hand every failing check and VALID finding to one fix worker that commits and pushes once, since each push starts a full CI run and cancels the one in flight. **Before any push to the open PR** (a fix or a rebase), auto-merge is turned off so the new head cannot merge on `ci-status` alone before the AI reviews re-run. The fix worker's brief carries the step, since the worker pushes: immediately before the push, and again right after it (someone may arm the old head while the worker runs), when `gh pr view --json autoMergeRequest --jq '.autoMergeRequest'` is non-null, run `gh pr merge --disable-auto`. Neither the worker nor the orchestrator re-arms, and no lane arms auto-merge. Verified 2026-09-26 against gh 2.98.0: [`gh pr merge`](https://cli.github.com/manual/gh_pr_merge) documents `--disable-auto`, and `gh pr view --help` lists the `autoMergeRequest` JSON field. Recheck when a gh release note renames either. Dispatch that fix worker as `implementation:implementer` when the `implementation` plugin is installed, its agent frontmatter carries the strong-tier model binding, so the fix lands on the strong tier, never below the original (when the original ran at the frontier tier, pass the frontier alias as a per-invocation `model`); when that plugin is absent, pass an explicit per-invocation `model` resolving the strong tier's current alias rather than letting the fix worker inherit the orchestrator's model. -1. **Post-green review pass, then hand off.** After CI is green, run one review pass. The fetch-once → validate → classify → threaded-reply → react → resolve-bot-thread loop is owned by `/source-control:pull-request`; this skill adds only the sequencing and the work-item linkage: fix findings via the same **fresh-subagent-into-the-persisted-worktree** re-dispatch the orchestrator-owned PR step defines (the orchestrator still never edits source). **Fix, don't file:** a small or medium VALID finding is fixed, never filed, and where an unrelated one lands is the scope test `/source-control:pull-request` owns. Only a finding that already reproduces on the base branch may defer; one this change introduced is fixed now at any size (the provenance test `/source-control:pull-request` owns). Of those, only three kinds are deferred: a **structural** finding, large enough to need its own planning pass; an **urgent, real** problem that cannot land in this PR; and a finding **blocked on research** this lane is not positioned to do. Nothing else is filed: no nits, no speculative items, no busy work. A **deferred finding requires a filed follow-up issue**. File it by invoking `/work-items:track add` via the Skill tool following the shared self-observation contract ([`${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md`](${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md): dedupe → categorize → fixed shape → `needs-triage`), then cite that issue **both** in the classification reply **and** in the PR's `## Related` section. **ensure the section exists first**: `/source-control:pull-request`'s scaffold carries `## Related` only when the repo's `pr_body_required_sections` requires it or a genuine reference already populated it at create time, so a deferred finding is frequently the first content that section ever holds. Adding it is a **read-modify-write**, never a bare `--body` replacement. `gh pr edit --body`/`--body-file` REPLACES the whole body (the same identity note the GitHub adapter's [PR closing-keyword mechanics](../../tools/work-item-tracker/adapters/github/README.md) documents for its own body edit), so read the current body first (`gh pr view --json body --jq '.body'`), append the `## Related` section (or its content, if the section already exists) to that read, and write the combined result back via `--body-file -`; a bare `gh pr edit --body "## Related\n..."` would silently drop the closing line and every section `/source-control:pull-request create` wrote. Make that body write before the push that carries the round's fixes, or, with no fix to push, while no CI run is in flight on the head; never as its own edit right after a push, since a body edit re-runs every workflow that triggers on `edited` beside the push's own run. A deferred finding cannot be resolved without it. Then hand the PR off by invoking `/source-control:babysit-prs` via the Skill tool (fleet loop, owned there). +1. **Post-green review pass, then hand off.** After CI is green, run one review pass. The fetch-once → validate → classify → threaded-reply → react → resolve-bot-thread loop is owned by `/source-control:pull-request`; this skill adds only the sequencing and the work-item linkage: fix findings via the same **fresh-subagent-into-the-persisted-worktree** re-dispatch the orchestrator-owned PR step defines (the orchestrator still never edits source). **Fix, don't file:** a small or medium VALID finding is fixed, never filed, and where an unrelated one lands is the scope test `/source-control:pull-request` owns. Only a finding that already reproduces on the base branch may defer; one this change introduced is fixed now at any size (the provenance test `/source-control:pull-request` owns). Of those, only three kinds are deferred: a **structural** finding, large enough to need its own planning pass; an **urgent, real** problem that cannot land in this PR; and a finding **blocked on research** this lane is not positioned to do. Nothing else is filed: no nits, no speculative items, no busy work. A **deferred finding requires a filed follow-up issue**. File it by invoking `/work-items:track add` via the Skill tool following the shared self-observation contract ([`${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md`](${CLAUDE_PLUGIN_ROOT}/reference/dogfood-filing.md): dedupe → categorize → fixed shape → `needs-triage`; per-user `ai_disclaimer`: `${user_config.ai_disclaimer}`), then cite that issue **both** in the classification reply **and** in the PR's `## Related` section. **ensure the section exists first**: `/source-control:pull-request`'s scaffold carries `## Related` only when the repo's `pr_body_required_sections` requires it or a genuine reference already populated it at create time, so a deferred finding is frequently the first content that section ever holds. Adding it is a **read-modify-write**, never a bare `--body` replacement. `gh pr edit --body`/`--body-file` REPLACES the whole body (the same identity note the GitHub adapter's [PR closing-keyword mechanics](../../tools/work-item-tracker/adapters/github/README.md) documents for its own body edit), so read the current body first (`gh pr view --json body --jq '.body'`), append the `## Related` section (or its content, if the section already exists) to that read, and write the combined result back via `--body-file -`; a bare `gh pr edit --body "## Related\n..."` would silently drop the closing line and every section `/source-control:pull-request create` wrote. Make that body write before the push that carries the round's fixes, or, with no fix to push, while no CI run is in flight on the head; never as its own edit right after a push, since a body edit re-runs every workflow that triggers on `edited` beside the push's own run. A deferred finding cannot be resolved without it. Then hand the PR off by invoking `/source-control:babysit-prs` via the Skill tool (fleet loop, owned there). 1. **Never-merge boundary.** This skill's lane ends at PR creation and the handoff above (review pass, then babysit). **Merging is the babysit lane or a human, never `work`**. Consistent with `/source-control:babysit-prs`'s safe default never merging (its opt-in `worker`/`autopilot` tiers merge only behind a deterministic readiness gate) and `/source-control:pull-request` merges being human-gated. The worker's worktree **persists** through the whole PR lifecycle, the same-worktree fix re-dispatch above depends on it, and is cleaned up only by whoever merges (`/source-control:pull-request merge` Phase 4, the babysit worker tier, or a human via `/source-control:worktree cleanup`), never mid-lifecycle and never by this lane. diff --git a/prompts/loops/loop-lane-profile-claude-code-plugins.md b/prompts/loops/loop-lane-profile-claude-code-plugins.md index 241754722e2..43e75f8e752 100644 --- a/prompts/loops/loop-lane-profile-claude-code-plugins.md +++ b/prompts/loops/loop-lane-profile-claude-code-plugins.md @@ -151,8 +151,9 @@ machines; neither on the attended box. > the direction that `/work-items:triage`'s mutation gate and the > self-observation filing contract require: triage, classify, label, > comment, file follow-up items, claim items, author branches and PRs, -> all without a human turn. Prefix every comment and item you create with -> the AI disclaimer specified by triage. You never merge. +> all without a human turn. Prefix comments and items you create with the +> AI disclaimer only as work-items' `ai_disclaimer` setting says. You never +> merge. > > **Discipline.** Invoke `/discipline:sweep-all`'s **full batch pass once per > cycle, at the cycle root only**, never from inside that skill's own audit diff --git a/prompts/loops/loop-lane-prompts.md b/prompts/loops/loop-lane-prompts.md index 0fdafac193c..7115b3daa30 100644 --- a/prompts/loops/loop-lane-prompts.md +++ b/prompts/loops/loop-lane-prompts.md @@ -520,8 +520,9 @@ variable that sets it). > the direction that `/work-items:triage`'s mutation gate and the > self-observation filing contract require: triage, classify, label, > comment, file follow-up items, claim items, author branches and PRs, -> all without a human turn. Prefix every comment and item you create with -> the AI disclaimer specified by triage. You never merge. +> all without a human turn. Prefix comments and items you create with the +> AI disclaimer only as work-items' `ai_disclaimer` setting says. You never +> merge. > > **Discipline.** Invoke `/discipline:sweep-all`'s **full batch pass once per > cycle, at the cycle root only**, never from inside that skill's own audit diff --git a/scripts/affected-tests-no-suite.txt b/scripts/affected-tests-no-suite.txt index 73efeb0ec39..a554e2eb133 100644 --- a/scripts/affected-tests-no-suite.txt +++ b/scripts/affected-tests-no-suite.txt @@ -205,3 +205,4 @@ docs/conventions/code-metrics.yaml docs/conventions/pr-pipeline.yaml docs/conventions/source-control.yaml docs/conventions/testing.yaml +docs/conventions/work-items.yaml From 25d2ee1973225bb131908899d9688a048b9fe93a Mon Sep 17 00:00:00 2001 From: "melodic-automation-lanes[bot]" <337595936+melodic-automation-lanes[bot]@users.noreply.github.com> Date: Mon, 26 Oct 2026 09:35:30 +0000 Subject: [PATCH 2/5] Address review feedback on work-resolve-issue/issue-8650-40284772685-1 Lane-Activity: pr-unblock/address-feedback --- .../feat-8650-ai-disclaimer-setting-5c2e9a41.md | 11 +++++++---- .../work-items/skills/work-loop/evals/evals.json | 15 ++++++++++++++- .../work-loop/reference/objection-window.md | 11 +++++++---- 3 files changed, 28 insertions(+), 9 deletions(-) diff --git a/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md b/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md index 2069128d30d..9a06202e2c0 100644 --- a/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md +++ b/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md @@ -13,8 +13,11 @@ bump: minor - **Lanes no longer prefix the AI disclaimer by default.** Triage, the work-loop intake sweep, attend-queue answers and autonomous filing from `work` and `scan-todos` add it only when - `ai_disclaimer` resolves `true`; set it to keep the old behavior. + `ai_disclaimer` resolves `true`; set it to keep the old behavior. A lane running on a person's + own `gh` token, with no `lane_app_slug`, now posts model-written items and comments under that + person's name with no AI marker at all; set `ai_disclaimer: true` there. - **The work-loop objection window no longer treats an escalation's answer as a new objection.** A - comment posted after a work-loop escalation marker and before the autonomous-eligible role label - was last applied answers that escalation, so an attend-queue answer without the disclaimer does - not re-escalate the item. + comment answers the newest work-loop escalation marker only when that marker is newer than the + window comment, the comment comes before the autonomous-eligible role label was last applied, + and its author is the trusted labeler who applied it. An attend-queue answer without the + disclaimer does not re-escalate the item; any other comment in that span is still an objection. diff --git a/plugins/work-items/skills/work-loop/evals/evals.json b/plugins/work-items/skills/work-loop/evals/evals.json index 9a069745ccf..11fde98d2c1 100644 --- a/plugins/work-items/skills/work-loop/evals/evals.json +++ b/plugins/work-items/skills/work-loop/evals/evals.json @@ -481,13 +481,26 @@ "id": 34, "name": "work-loop-escalation-answer-is-not-a-new-objection", "prompt": "/work-items:work-loop\n\nA cycle starts at 2026-10-05T12:00:00Z. ai_disclaimer resolves false. #306 is agent-ready and work-class: mechanical with no open closing PR. Its objection window comment from the seam's write identity ended at 2026-10-04T10:00:00Z. At 2026-10-04T11:00:00Z the write identity posted a lane=work-loop kind=escalated marker for a maintainer's objection. At 2026-10-05T09:00:00Z the operator's account (write permission, not the write identity, no AI disclaimer) commented 'Decision: keep the old config format.' At 09:01 the same operator applied agent-ready; label-provenance reports that application as trusted. No comment follows it.", - "expected_output": "The operator's 09:00 comment sits after the work-loop escalation marker and before the latest trusted application of agent-ready, so it answers that escalation and is not a new objection; the maintainer's earlier objection is answered by the marker. With the window over, #306 continues to classification and is not escalated again.", + "expected_output": "The operator's 09:00 comment sits after the newest work-loop escalation marker, which is newer than the window comment, and before the latest trusted application of agent-ready, and its author is the actor of that application, so it answers that escalation and is not a new objection; the maintainer's earlier objection is answered by the marker. With the window over, #306 continues to classification and is not escalated again.", "files": [], "expectations": [ "Does not treat the operator's 09:00 comment as a new objection", "Does not escalate #306 again", "Continues #306 to classification" ] + }, + { + "id": 35, + "name": "work-loop-other-persons-comment-before-relabel-is-still-an-objection", + "prompt": "/work-items:work-loop\n\nA cycle starts at 2026-10-05T12:00:00Z. ai_disclaimer resolves false. #307 is agent-ready and work-class: mechanical with no open closing PR. Its objection window comment from the seam's write identity ended at 2026-10-04T10:00:00Z. At 2026-10-04T11:00:00Z the write identity posted a lane=work-loop kind=escalated marker for a maintainer's objection. At 2026-10-05T08:30:00Z an outside commenter (no write permission, not a bot, no AI disclaimer) commented 'I object, this breaks the importer.' At 09:01 the operator applied agent-ready; label-provenance reports that application as trusted, with the operator as its actor. No comment follows it.", + "expected_output": "The outside commenter's 08:30 comment sits between the escalation marker and the latest agent-ready application, but its author is not that application's actor, so it does not answer the escalation: it is a new human objection. #307 is escalated as kind=escalated with a one-line question naming that comment, leaves the autonomous frontier, and the cycle report lists objection: #307 -> escalated.", + "files": [], + "expectations": [ + "Treats the outside commenter's 08:30 comment as a new objection, since its author is not the actor of the trusted agent-ready application", + "Escalates #307 as kind=escalated naming the objecting comment", + "Does not continue #307 to classification", + "Lists objection: #307 -> escalated in the cycle report" + ] } ] } diff --git a/plugins/work-items/skills/work-loop/reference/objection-window.md b/plugins/work-items/skills/work-loop/reference/objection-window.md index 7fefa7769be..3ffc7c324cf 100644 --- a/plugins/work-items/skills/work-loop/reference/objection-window.md +++ b/plugins/work-items/skills/work-loop/reference/objection-window.md @@ -18,10 +18,13 @@ a window. Check in this order: marker answers an objection only when the seam's configured write identity posted it, as with the window marker; the same marker text from any other author is an ordinary comment, so a pasted marker cannot cancel an objection (and, posted by a person, is itself an objection). - A comment posted after such a marker and before the latest application of the - autonomous-eligible role label (its `at` in the labeler precondition's `label-provenance` read, - from a trusted labeler) answers that escalation, as `/work-items:attend-queue` writes its - answers, and is not a new objection. A person here is an + A comment answers that escalation, as `/work-items:attend-queue` writes its answers, and is not + a new objection, only when all of these hold: it was posted after the newest such marker, and + that marker is newer than the window comment; it was posted before the latest application of + the autonomous-eligible role label (its `at` in the labeler precondition's `label-provenance` + read), which that read reports as trusted; and its author is that application's `actor`. Any + other comment in that span, from another person or after an older marker, is an objection. A + person here is an author that is not the seam's write identity, is not a bot account (the tracker reports its user type as `Bot`, or its login ends in `[bot]`), and whose comment carries no AI disclaimer (a lane adds one only when the setting in From a74553791a8b2f4230a9c312e6a11096c67bf471 Mon Sep 17 00:00:00 2001 From: "melodic-automation-lanes[bot]" <337595936+melodic-automation-lanes[bot]@users.noreply.github.com> Date: Mon, 26 Oct 2026 09:55:29 +0000 Subject: [PATCH 3/5] Fix failing CI checks on work-resolve-issue/issue-8650-40284772685-1 Lane-Activity: pr-unblock/fix-ci --- plugins/work-items/skills/work-loop/SKILL.md | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/plugins/work-items/skills/work-loop/SKILL.md b/plugins/work-items/skills/work-loop/SKILL.md index 47ed528fa50..5a1e01e2fe1 100644 --- a/plugins/work-items/skills/work-loop/SKILL.md +++ b/plugins/work-items/skills/work-loop/SKILL.md @@ -82,8 +82,7 @@ owned by [reference/telemetry-upsert.md](reference/telemetry-upsert.md). Compose invoke the script; do not write the comment by hand. When the bound provider is not `github`, this upsert is unavailable: carry the same telemetry -content, state block included, in the lane's cycle report/log, noting the comment surface is -absent. +content, state block included, in the lane's cycle report/log, noting the comment surface is absent. The comment carries the human-readable cycle report plus a machine-readable **durable loop state** block, re-read at every cycle start (conversation context is compaction-lossy, the comment is @@ -230,14 +229,13 @@ the former default-on pause). comment or item it creates carries the AI disclaimer when the setting resolves `true` ([`${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md`](${CLAUDE_PLUGIN_ROOT}/reference/ai-disclaimer.md); the per-user value it reads is `${user_config.ai_disclaimer}`). Sweep hardening: an advisory issue - authored by a workflow bot routes to the human-gated role label by default (this also lets drain - exits terminate against automated intake). Applied together with a machine-marked - `kind=routed-advisory` escalation comment (step 5's marker shape and record write), so the - routing surfaces in the attended queue's escalated view instead of vanishing behind a bare - label. When the launch prompt bars this lane from writing `work-class:` labels, a delegable or - decision-defaulted outcome takes triage's "Lane barred from recording a class" branch (human-gated - role plus a `kind=escalated` marker carrying the proposed class), and that marker carries - step 5's record write the same way. + authored by a workflow bot routes to the human-gated role label by default (this also lets drain exits + terminate against automated intake). Applied together with a machine-marked `kind=routed-advisory` + escalation comment (step 5's marker shape and record write), so the routing surfaces in the attended + queue's escalated view instead of vanishing behind a bare label. When the launch prompt bars this + lane from writing `work-class:` labels, a delegable or decision-defaulted outcome takes triage's + "Lane barred from recording a class" branch (human-gated role plus a `kind=escalated` marker carrying + the proposed class), and that marker carries step 5's record write the same way. 3. **Admission gate.** Drop every frontier candidate that is already in flight (the gate's in-flight precondition below) or held by its Objection window ([reference/objection-window.md](reference/objection-window.md), which also escalates an objected one), then classify the rest and admit per the gate. Fail-closed. From e9b4396849925d8500135c39fc3a35a5f9207828 Mon Sep 17 00:00:00 2001 From: "melodic-automation-lanes[bot]" <337595936+melodic-automation-lanes[bot]@users.noreply.github.com> Date: Tue, 27 Oct 2026 03:46:01 +0000 Subject: [PATCH 4/5] Address review feedback on work-resolve-issue/issue-8650-40284772685-1 Lane-Activity: pr-unblock/address-feedback --- .../feat-8650-ai-disclaimer-setting-5c2e9a41.md | 3 +++ plugins/work-items/.claude-plugin/plugin.json | 2 +- plugins/work-items/README.md | 2 +- plugins/work-items/reference/ai-disclaimer.md | 7 ++++--- .../work-items/skills/work-loop/evals/evals.json | 13 +++++++++++++ .../skills/work-loop/reference/objection-window.md | 9 +++++---- 6 files changed, 27 insertions(+), 9 deletions(-) diff --git a/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md b/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md index 9a06202e2c0..48202e06d8f 100644 --- a/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md +++ b/.changes/work-items/feat-8650-ai-disclaimer-setting-5c2e9a41.md @@ -21,3 +21,6 @@ bump: minor window comment, the comment comes before the autonomous-eligible role label was last applied, and its author is the trusted labeler who applied it. An attend-queue answer without the disclaimer does not re-escalate the item; any other comment in that span is still an objection. +- **The objection window no longer reads the AI disclaimer.** Who counts as a person rests on the + author alone (not the write identity, not a bot account): a comment carrying the prefix is still + a person's objection, since anyone can type it. diff --git a/plugins/work-items/.claude-plugin/plugin.json b/plugins/work-items/.claude-plugin/plugin.json index 3ddb900fec3..df32217bbd7 100644 --- a/plugins/work-items/.claude-plugin/plugin.json +++ b/plugins/work-items/.claude-plugin/plugin.json @@ -40,7 +40,7 @@ "ai_disclaimer": { "type": "boolean", "title": "AI disclaimer on lane posts", - "description": "When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false. The repository's docs/conventions/work-items.yaml overrides this per repo.", + "description": "When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false, a change on update: lanes always added it before this setting existed, so set true to keep that. The repository's docs/conventions/work-items.yaml overrides this per repo.", "default": false }, "work_dispatch_concurrency_cap": { diff --git a/plugins/work-items/README.md b/plugins/work-items/README.md index b40d9455c0b..a0dfa6a6a6f 100644 --- a/plugins/work-items/README.md +++ b/plugins/work-items/README.md @@ -231,7 +231,7 @@ reads it from. | `decompose_container_publish` | boolean | *(none)* | `CLAUDE_PLUGIN_OPTION_DECOMPOSE_CONTAINER_PUBLISH` | When true, /work-items:decompose pre-selects the spec-container offer in its approval round for multi-session breakdowns; the approval gate stays mandatory. Unset or false: the plain ask, with a default answer of no. No default is declared, so unset stays distinct from false. | | `triage_repro_count` | number
*min 1* | `2` | `CLAUDE_PLUGIN_OPTION_TRIAGE_REPRO_COUNT` | Independent reproductions /work-items:triage needs, each with evidence in its comment, before it marks a bug confirmed. Default 2. The repository's docs/conventions/work-items.yaml overrides this per repo. | | `triage_objection_window_hours` | number
*min 0* | `0` | `CLAUDE_PLUGIN_OPTION_TRIAGE_OBJECTION_WINDOW_HOURS` | Hours after triage makes an item autonomous-eligible during which the work-loop lane skips it, so a person can object in a reply. 0 (the default) is off. The repository's docs/conventions/work-items.yaml overrides this per repo. | -| `ai_disclaimer` | boolean | `false` | `CLAUDE_PLUGIN_OPTION_AI_DISCLAIMER` | When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false. The repository's docs/conventions/work-items.yaml overrides this per repo. | +| `ai_disclaimer` | boolean | `false` | `CLAUDE_PLUGIN_OPTION_AI_DISCLAIMER` | When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false, a change on update: lanes always added it before this setting existed, so set true to keep that. The repository's docs/conventions/work-items.yaml overrides this per repo. | | `work_dispatch_concurrency_cap` | number
*min 1* | *(none)* | `CLAUDE_PLUGIN_OPTION_WORK_DISPATCH_CONCURRENCY_CAP` | Maximum worker rows /work-items:work's autonomous execute step lets /implementation:implement-dispatch run at once in one plan phase, passed as its --wave-cap ceiling. A fractional value is floored. Unset: implement-dispatch's implement_dispatch_wave_cap option, else its internal default. | | `lane_instance` | string | *(none)* | `CLAUDE_PLUGIN_OPTION_LANE_INSTANCE` | Writer identity for this machine's work-loop telemetry, the suffix of its telemetry marker, so concurrent lane instances never overwrite each other. Absent: the sanitized lowercased hostname. Must match ^\[a-z0-9\]\[a-z0-9-\]{0,31}$. It appears verbatim in tracker comments. | | `lane_app_slug` | string | *(none)* | `CLAUDE_PLUGIN_OPTION_LANE_APP_SLUG` | Slug of the GitHub App whose installation token the work-loop lane runs under, passed as --app-slug to its telemetry upsert so the write identity resolves to \[bot\]. Needed only under an App token. Absent: the identity resolves from gh api user. Must match ^\[a-z0-9\]\[a-z0-9-\]*$. | diff --git a/plugins/work-items/reference/ai-disclaimer.md b/plugins/work-items/reference/ai-disclaimer.md index d3bedc9216f..f43c2efe5eb 100644 --- a/plugins/work-items/reference/ai-disclaimer.md +++ b/plugins/work-items/reference/ai-disclaimer.md @@ -45,6 +45,7 @@ condition and the sentence, rather than restating either: - [`../skills/work-loop/SKILL.md`](../skills/work-loop/SKILL.md): the intake sweep - Loop-lane standing prompts and lane-absorption skills: reconcile against this contract -A reader that tells lane text from a person's by the prefix (the work-loop objection window) cannot -rely on it while the setting resolves `false`, so it does not rest on the prefix alone: the -objection window also reads the author and when the item was returned to the frontier. +The prefix is provenance for readers, never evidence of who wrote a comment: it is absent while the +setting resolves `false`, and anyone can type it. A reader that must tell lane text from a person's +(the work-loop objection window) reads the author and when the item was returned to the frontier, +never the prefix. diff --git a/plugins/work-items/skills/work-loop/evals/evals.json b/plugins/work-items/skills/work-loop/evals/evals.json index 11fde98d2c1..0b6f09479c6 100644 --- a/plugins/work-items/skills/work-loop/evals/evals.json +++ b/plugins/work-items/skills/work-loop/evals/evals.json @@ -501,6 +501,19 @@ "Does not continue #307 to classification", "Lists objection: #307 -> escalated in the cycle report" ] + }, + { + "id": 36, + "name": "work-loop-disclaimer-prefixed-comment-is-still-an-objection", + "prompt": "/work-items:work-loop\n\nA cycle starts at 2026-10-05T12:00:00Z. ai_disclaimer resolves false. #308 is agent-ready and work-class: mechanical with no open closing PR. Its objection window comment from the seam's write identity ended at 2026-10-05T10:00:00Z. At 09:30 an outside commenter (not the write identity, user type User, login not ending in [bot]) commented: '*This was generated by AI during triage.*' on its first line, then 'I object, this breaks the importer.' No work-loop escalation marker is on the item.", + "expected_output": "The 09:30 comment's author is not the seam's write identity and not a bot account, so it is a person's comment whatever its text says; the AI disclaimer line in it is not evidence of a lane author. It is a human objection: #308 is escalated as kind=escalated with a one-line question naming that comment, leaves the autonomous frontier, and the cycle report lists objection: #308 -> escalated.", + "files": [], + "expectations": [ + "Treats the 09:30 comment as a person's objection despite its AI disclaimer line", + "Escalates #308 as kind=escalated naming the objecting comment", + "Does not continue #308 to classification", + "Lists objection: #308 -> escalated in the cycle report" + ] } ] } diff --git a/plugins/work-items/skills/work-loop/reference/objection-window.md b/plugins/work-items/skills/work-loop/reference/objection-window.md index 3ffc7c324cf..5867092d5e7 100644 --- a/plugins/work-items/skills/work-loop/reference/objection-window.md +++ b/plugins/work-items/skills/work-loop/reference/objection-window.md @@ -25,10 +25,11 @@ a window. Check in this order: read), which that read reports as trusted; and its author is that application's `actor`. Any other comment in that span, from another person or after an older marker, is an objection. A person here is an - author that is not the seam's write identity, is not a bot account (the tracker reports its - user type as `Bot`, or its login ends in `[bot]`), and whose comment carries no AI disclaimer - (a lane adds one only when the setting in - [`../../../reference/ai-disclaimer.md`](../../../reference/ai-disclaimer.md) resolves `true`). Step 5 escalates the item as + author that is not the seam's write identity and is not a bot account (the tracker reports its + user type as `Bot`, or its login ends in `[bot]`). What a comment says never makes it not a + person's: an AI disclaimer in it + ([`../../../reference/ai-disclaimer.md`](../../../reference/ai-disclaimer.md)) is text anyone + can type, so it is never evidence of a lane author. Step 5 escalates the item as `kind=escalated`, whose one-line question names the objecting comment, and it leaves the autonomous frontier. The cycle report lists it as `objection: # -> escalated`. 2. **Window still open** (`` after the cycle-start time): skip the candidate this cycle. It is From 76c5ff5fbef7a9610546ab38d41b76baecc8a784 Mon Sep 17 00:00:00 2001 From: "melodic-automation-lanes[bot]" <337595936+melodic-automation-lanes[bot]@users.noreply.github.com> Date: Tue, 27 Oct 2026 03:58:34 +0000 Subject: [PATCH 5/5] Fix failing CI checks on work-resolve-issue/issue-8650-40284772685-1 Lane-Activity: pr-unblock/fix-ci --- plugins/work-items/.claude-plugin/plugin.json | 2 +- plugins/work-items/README.md | 5 ++++- 2 files changed, 5 insertions(+), 2 deletions(-) diff --git a/plugins/work-items/.claude-plugin/plugin.json b/plugins/work-items/.claude-plugin/plugin.json index df32217bbd7..d76c6f0ee9a 100644 --- a/plugins/work-items/.claude-plugin/plugin.json +++ b/plugins/work-items/.claude-plugin/plugin.json @@ -40,7 +40,7 @@ "ai_disclaimer": { "type": "boolean", "title": "AI disclaimer on lane posts", - "description": "When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false, a change on update: lanes always added it before this setting existed, so set true to keep that. The repository's docs/conventions/work-items.yaml overrides this per repo.", + "description": "When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false, so lanes omit it. The repository's docs/conventions/work-items.yaml overrides this per repo.", "default": false }, "work_dispatch_concurrency_cap": { diff --git a/plugins/work-items/README.md b/plugins/work-items/README.md index a0dfa6a6a6f..090e6230b7f 100644 --- a/plugins/work-items/README.md +++ b/plugins/work-items/README.md @@ -218,6 +218,9 @@ from being overwritten by another's. **`work_loop_frontier_item_cap_ceiling`.** A quota guard: items carrying `capability-tier: frontier` are bounded by this ceiling instead of the general one. +**`ai_disclaimer`.** The false default is a change on update: lanes always added the disclaimer +before this setting existed, so set it true to keep that behavior. + ### Options reference @@ -231,7 +234,7 @@ reads it from. | `decompose_container_publish` | boolean | *(none)* | `CLAUDE_PLUGIN_OPTION_DECOMPOSE_CONTAINER_PUBLISH` | When true, /work-items:decompose pre-selects the spec-container offer in its approval round for multi-session breakdowns; the approval gate stays mandatory. Unset or false: the plain ask, with a default answer of no. No default is declared, so unset stays distinct from false. | | `triage_repro_count` | number
*min 1* | `2` | `CLAUDE_PLUGIN_OPTION_TRIAGE_REPRO_COUNT` | Independent reproductions /work-items:triage needs, each with evidence in its comment, before it marks a bug confirmed. Default 2. The repository's docs/conventions/work-items.yaml overrides this per repo. | | `triage_objection_window_hours` | number
*min 0* | `0` | `CLAUDE_PLUGIN_OPTION_TRIAGE_OBJECTION_WINDOW_HOURS` | Hours after triage makes an item autonomous-eligible during which the work-loop lane skips it, so a person can object in a reply. 0 (the default) is off. The repository's docs/conventions/work-items.yaml overrides this per repo. | -| `ai_disclaimer` | boolean | `false` | `CLAUDE_PLUGIN_OPTION_AI_DISCLAIMER` | When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false, a change on update: lanes always added it before this setting existed, so set true to keep that. The repository's docs/conventions/work-items.yaml overrides this per repo. | +| `ai_disclaimer` | boolean | `false` | `CLAUDE_PLUGIN_OPTION_AI_DISCLAIMER` | When true, autonomous work-items lanes (triage, work-loop, attend-queue answers, autonomous filing) start each tracker item and comment they create with the AI disclaimer. Default false, so lanes omit it. The repository's docs/conventions/work-items.yaml overrides this per repo. | | `work_dispatch_concurrency_cap` | number
*min 1* | *(none)* | `CLAUDE_PLUGIN_OPTION_WORK_DISPATCH_CONCURRENCY_CAP` | Maximum worker rows /work-items:work's autonomous execute step lets /implementation:implement-dispatch run at once in one plan phase, passed as its --wave-cap ceiling. A fractional value is floored. Unset: implement-dispatch's implement_dispatch_wave_cap option, else its internal default. | | `lane_instance` | string | *(none)* | `CLAUDE_PLUGIN_OPTION_LANE_INSTANCE` | Writer identity for this machine's work-loop telemetry, the suffix of its telemetry marker, so concurrent lane instances never overwrite each other. Absent: the sanitized lowercased hostname. Must match ^\[a-z0-9\]\[a-z0-9-\]{0,31}$. It appears verbatim in tracker comments. | | `lane_app_slug` | string | *(none)* | `CLAUDE_PLUGIN_OPTION_LANE_APP_SLUG` | Slug of the GitHub App whose installation token the work-loop lane runs under, passed as --app-slug to its telemetry upsert so the write identity resolves to \[bot\]. Needed only under an App token. Absent: the identity resolves from gh api user. Must match ^\[a-z0-9\]\[a-z0-9-\]*$. |