Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions .changes/implementation/fix-6916-model-pin-claims-a4c19e27.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,7 @@
---
bump: patch
---

### Fixed

- `/implementation:implement-dispatch` and the implementer agents no longer say a frontmatter model or effort pin always holds: each now names the upstream section listing what ranks over the pin.
11 changes: 7 additions & 4 deletions plugins/implementation/agents/implementer.md
Original file line number Diff line number Diff line change
Expand Up @@ -153,13 +153,16 @@ bookkeeping must never lower it for the phase implementation too. The binding is
model-config row the pointer below names: a phase brief is scoped, day-to-day engineering work, and
the phase verifier that checks it runs at `high`.

- **Pointer:** the `medium` row of
- **Pointer:** what each alias resolves to, per provider, in
[model config: model aliases](https://code.claude.com/docs/en/model-config#model-aliases);
the `medium` row of
[model config: choose an effort level](https://code.claude.com/docs/en/model-config#choose-an-effort-level);
[optimizing for cost and intelligence: compare models on cost per task](https://platform.claude.com/docs/en/about-claude/models/optimizing-for-cost-and-intelligence#compare-models-on-cost-per-task);
what effort a subagent runs at without the pin, in the `effort` row of
[sub-agents: supported frontmatter fields](https://code.claude.com/docs/en/sub-agents#supported-frontmatter-fields),
and that plus the Agent tool's per-call `effort` and its rank over this pin, in
[sub-agents: choose an effort level](https://code.claude.com/docs/en/sub-agents#choose-an-effort-level).
- **As of:** 2026-10-11.
- **Recheck trigger:** next model release, or the subagents `effort` row or effort section changes
what a subagent's effort follows, the precedence, or which spawns honor a per-call `effort`.
- **As of:** 2026-10-31.
- **Recheck trigger:** next model release, the aliases section changes how an alias resolves, or
the subagents `effort` row or effort section changes what a subagent's effort follows, the
precedence, or which spawns honor a per-call `effort`.
10 changes: 5 additions & 5 deletions plugins/implementation/agents/scoped-implementer.md
Original file line number Diff line number Diff line change
Expand Up @@ -142,7 +142,7 @@ decision, a cross-module contract change, or a file outside the fence, that is t
brief's escalation clause names: STOP and report, so the orchestrator can re-dispatch the phase to
`implementation:implementer`. Never stretch to finish it here.

`effort` is pinned alongside `model` so the phase never runs at the session's level. A dispatcher
`effort` is pinned alongside `model` so the phase does not follow the session's level. A dispatcher
passes no `effort` here: neither the Agent tool's per-call parameter nor a Workflow `agent()`
call's.

Expand All @@ -154,11 +154,11 @@ the effort-pin owner's ruling of 2026-10-01.
- **Pointer**: for the model split, see
<https://code.claude.com/docs/en/costs#choose-the-right-model>; for the effort levels, see
<https://code.claude.com/docs/en/model-config#adjust-effort-level> and
<https://platform.claude.com/docs/en/build-with-claude/effort#recommended-effort-levels-for-claude-sonnet-5-5>;
<https://platform.claude.com/docs/en/build-with-claude/effort#recommended-effort-levels>;
for the subagent model order and allowlist substitution, see
<https://code.claude.com/docs/en/sub-agents#choose-a-model>; for the per-spawn `effort` and its
rank over the pin, see <https://code.claude.com/docs/en/sub-agents#choose-an-effort-level>.
- **As of**: 2026-10-11
<https://code.claude.com/docs/en/sub-agents#choose-a-model>; for the per-spawn `effort` and
every source that ranks over the pin, see <https://code.claude.com/docs/en/sub-agents#choose-an-effort-level>.
- **As of**: 2026-10-31
- **Recheck trigger**: the subagents effort section changes precedence or which spawns honor
`effort`, the costs section changes its model split, either effort-levels section changes its
advice for `medium`, the allowlist substitution changes, or the `sonnet` alias moves to a new
Expand Down
2 changes: 1 addition & 1 deletion plugins/implementation/skills/implement-dispatch/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -109,7 +109,7 @@ Because the orchestrator stays on the default branch, **every source-touching op

**Do not dispatch a brief with a gap.** When the plan, the Brief and the design give no answer for an item above and none can be derived, hold that row back, never sending it with the item blank or guessed. Report the unit as unscoped, naming each item you could not fill, and route it as a divergence (step 5). An item that does not apply, such as item 9 with no dedicated worktree or item 12 as `Design: none`, is not missing.
2. **Dispatch** each worker by its phase's routing row, the `Model` column. Never rely on root inheritance, and never dispatch source-editing work through a generic subagent type, a no-commit plan included. A worker runs where this session runs; this skill starts no worker on another host (cloud placement is chosen per work item before this skill starts, by the execution-target contract, `docs/contracts/execution-target/README.md` in this plugin's marketplace repository).
- **No row, no `Model` value, `opus`, or any value not listed here**: `implementation:implementer`. Its `model` frontmatter binds the strong tier's current alias, so an unqualified dispatch lands on the intended tier regardless of the orchestrator's model.
- **No row, no `Model` value, `opus`, or any value not listed here**: `implementation:implementer`. Its `model` frontmatter binds the strong tier's current alias, so an unqualified dispatch needs no per-call `model`. For how the harness resolves that binding, see [sub-agents: choose a model](https://code.claude.com/docs/en/sub-agents#choose-a-model) and [run every subagent on one model](https://code.claude.com/docs/en/sub-agents#run-every-subagent-on-one-model) (as of 2026-10-31; recheck when either section changes what overrides a frontmatter `model`).
- **`sonnet`**: first read the provider with `printenv | grep -E '^(CLAUDE_CODE_USE_(BEDROCK|VERTEX|FOUNDRY|ANTHROPIC_AWS|MANTLE)|ANTHROPIC_DEFAULT_SONNET_MODEL)='`. When a `CLAUDE_CODE_USE_BEDROCK`, `CLAUDE_CODE_USE_VERTEX`, `CLAUDE_CODE_USE_FOUNDRY`, `CLAUDE_CODE_USE_ANTHROPIC_AWS` or `CLAUDE_CODE_USE_MANTLE` value is non-empty and `ANTHROPIC_DEFAULT_SONNET_MODEL` is unset or empty, the unpinned `sonnet` alias is not trusted on that provider: dispatch `implementation:implementer` instead. Otherwise dispatch `implementation:scoped-implementer` with an explicit per-invocation `model: sonnet`, so a caller's standing per-spawn model rule (a "pass `model: opus` on every spawn" rule) cannot apply. Downward routing happens only this way.
- **`frontier`, or a security-surface work class whatever the row says**: `implementation:implementer` at the frontier tier's current alias, per the next rule.

Expand Down
Loading