You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
source-control(babysit-prs): absorb v4 loop-prompt operating rules into the skill #477
Category: behavioral/operational · Source: 2026-07-19 deep-dive audit (external + lane self-audit). Session-scoped loop-prompt rules that are really skill gaps — each should live in the skill so bare invocations get them without a hand-authored prompt.
Absorb into the skill
Review-response ownership: pipeline/lane-authored open PRs — fetch threads, fix branch-owned findings, classification-table reply, then RESOLVE each addressed thread (or explicit defer + follow-up issue #). Audit found 12/12 findings addressed but 0/~10 threads resolved.
This was generated by AI during triage. · Category: behavioral/operational · Priority:priority: high (week directive — pipeline improving itself). Routed needs-human + wayfind: design.
Plan: absorb v4 loop-prompt operating rules into babysit-prs
Triage of the six rules against the current skill. Two are clean (one already fully enforced, one a mechanical addition); one shared "defer → follow-up issue" mechanism is planned once and reused; two defer to existing issues; two (bullets 1 and 3) are genuinely open design and gate this plan on a maintainer decision.
A cross-cutting mechanism used by bullets 1 and 3 — on any decide-and-defer, file a tracked follow-up issue and post its number in the classification-table reply — is specified once here and referenced by both; do not plan it twice.
Per-bullet disposition
1 — Review-response ownership (fetch → fix → classification reply → RESOLVE addressed threads, or defer + follow-up issue #) — needs human. Not a simple gap: the skill contradicts itself. The authoritative tier policy states four times that the safe/default tier never resolves threads (description, purpose, per-action autonomy table, cross-tier invariants), yet the safe-tier loop's own NEVER-do list mandates "never skip resolving a BOT-authored thread" (author-conditional, after fix+reply). The classification-table reply itself is already fully specified; thread resolution via source-control-babysit-resolve-thread exists but is gated to worker (pre-push-outdated bot threads) and autopilot (--include-human). The rule wants addressed bot/AI-review threads on your own pipeline/lane-authored PRs closed by the default invocation. Whichever direction the maintainer picks, one of the two contradicting statements must change — this is a product-intent call about the default tier's safety posture, not a mechanical absorption.
Acceptance: SKILL.md and the safe-tier loop agree on exactly which threads the default tier resolves on your own PRs; each addressed thread ends resolved (verified via the wrapper's per-thread JSON action, not exit code) or carries an explicit defer + follow-up issue #; human threads still never auto-resolved outside autopilot.
2 — Never re-trigger a review for an unchanged head SHA — already covered. The review-trigger interface (request_review.py) enforces one-shot-per-head: it refuses a post when a durable request/attempt record exists for the current head SHA, and a new observation window opens only on a genuinely new SHA. Dormant (and thus vacuously safe) until the review-trigger config slots are set. No change beyond optionally surfacing the invariant in the cross-tier list for visibility.
Acceptance: no behavioral change required; invariant holds whenever the module is active.
3 — Fix-round cap per head (3) → decide-and-defer + tracked follow-up issue — needs human. Three cap concepts must be reconciled: (a) the existing "max 3 CI fix iterations per pass", (b) the advisory backstop (~high default, ledger-enforced via manage_feedback_ledger.py record-advisory-round, deliberately not a normal limit), and (c) the proposed 3-per-head cap. The decision that makes this open: does the 3-per-head cap apply to blocking defects — reversing the deliberate, repeated invariant that blocking defects (failing CI, P0/P1, regressions) are never capped — or advisory only? No safe default exists; the two readings encode opposite convergence philosophies. Overlaps babysit-prs: codify the decide-and-defer non-convergence resolution playbook (own-once, no duplication of safety.md) #375 (non-convergence) — reconcile there, do not double-plan. The defer outcome reuses the shared follow-up-issue mechanism above.
Acceptance: one documented cap keyed to head SHA with an explicit blocking-vs-advisory scope; on cap, a decide-and-defer that files a follow-up issue and reports the number; no silent contradiction with the "blocking never capped" invariant.
4 — Merge-gate mode as a documented tier (hard STOP + MERGE-READY: [list] terminal output) — defer to feat(source-control): babysit autopilot merge tier — bot-account review + criteria-gated merge #476. New tier surface; the MERGE-READY: [list] per-cycle terminal-output convention and hard STOP do not exist today (only the source-control-babysit-merge wrapper and the worker/autopilot tiers). Planned under the merge-tier issue; not re-planned here.
5 — Parallel fan-out self-observation timing ("file after any ONE worker returns") — absorb (mechanical). Add an incremental reporting-cadence rule to the orchestration contract: the main agent surfaces each material observation as soon as any single worker returns, rather than withholding the report until the whole batch integrates. Scoped strictly to reporting timing — it must not alter batch dispatch/concurrency ordering (batch → integrate → next batch stays intact).
Acceptance: orchestration states that per-worker material outcomes are reported on first return within a cycle; concurrency-cap batching semantics unchanged.
Needs human because: bullets 1 and 3 are open design with no safe default. Bullet 1 forces a resolution of an existing SKILL.md-vs-safe-loop contradiction and decides whether the default tier resolves threads (a headline safety-posture change). Bullet 3 requires deciding whether a 3-per-head cap overrides the deliberate "blocking defects are never capped" invariant. Both are product-intent calls a maintainer must make before implementation; bullets 2/5 (already-covered / mechanical) and 4/6 (defer to #476 / #472) can proceed independently once 1 and 3 are decided.
Elevation (2026-07-19 merge pass): thread resolution is not just hygiene — the branch ruleset has required_review_thread_resolution, so unresolved (even fully-addressed) threads are MERGE-BLOCKING. Six digest-cleared PRs refused to merge until 9 addressed threads were manually resolved. The resolve-after-addressing checklist item in this umbrella is pipeline-critical; until it lands in the skill, the v5 lane rules carry it.
This was generated by AI (needs-human decision-brief pass).
Decision brief (2-minute read)
Problem. The triage plan cleared four of the six loop-prompt rules (two done/mechanical, bullets 4 and 6 deferred to #476/#472) but left bullets 1 and 3 open with no recommendation. This brief recommends a resolution for each; neither is settled elsewhere.
Today. The rules already run in the lane prompt, not the skill: .work/leave-week-priorities/loop-prompts/terminal-1-babysit.md (v7) rule 3 (fetch → fix → classification reply → RESOLVE each addressed thread, or defer + follow-up issue #), rule 6b (never re-trigger an unchanged head SHA; cap fix-rounds at 3/head → decide-and-defer #375), rule 6c (MERGE-READY line), rule 9 (file after any one worker returns). Skill home: plugins/source-control/skills/babysit-prs/. The contradiction the triage names is real — SKILL says the safe/default tier never resolves threads, while the safe-tier loop's NEVER-list mandates "never skip resolving a BOT-authored thread"; the resolver exists (bin/source-control-babysit-resolve-thread, scripts/babysit_resolve_thread.py) but is gated to worker/autopilot.
Decisive fact (already in this issue's elevation comment — build on it, not restated). The branch ruleset carries required_review_thread_resolution, so addressed-but-unresolved threads are merge-blocking; six digest-cleared PRs refused to merge until threads were resolved by hand. That breaks the bullet-1 tie.
Bullet 1 — does the default tier resolve threads?
RECOMMENDED — yes, bounded to its OWN addressed threads. On pipeline/lane-authored PRs, after fix + classification reply, the default tier resolves each addressed BOT/AI-review thread; human-authored threads still never auto-resolve outside autopilot; PRs you don't own are untouched. Rationale: the merge-blocking ruleset puts resolution on the merge path, and the narrow scope (own PRs, addressed, non-human) keeps the safety-posture change contained. Resolve the SKILL-vs-loop contradiction by editing SKILL to carve this bounded exception. Reversible (policy/doc edit; resolver already exists).
Keep default tier never-resolves (status quo). Preserves the blanket invariant but leaves every pipeline PR merge-blocked until a higher tier or a human resolves — the exact failure the elevation documents. Reversible, but re-incurs the manual toil.
Bullet 3 — does the 3-per-head cap apply to blocking defects?
T2 routing plan — decided-state implementation map
Re-intake pass after the operator acceptance (2026-07-19T22:35, decision-session batch 2): the RECOMMENDED options in the decision brief were accepted verbatim, which closes the only two open design calls (bullets 1 and 3). This comment does not re-derive that rationale — see the decision brief above — it consolidates the now-decided state into an implementation map and routes.
North-star note (sort, not tier): this is a pipeline self-improvement item and is relevant to the week's throughput north-star, so it should sort early within its tier — but per the #552 rebaseline that alone does not earn priority: high (not blocking other work, not scaling-chain). Routed priority: medium.
Per-bullet disposition (skill home · behavioral contract · acceptance)
1 — Review-response ownership → RESOLVE addressed threads — decided (absorb + fix contradiction). Operator ACCEPTED: on pipeline/lane-authored PRs the default tier, after fix + classification-table reply, resolves each addressed bot/AI-review thread it owns; human threads still never auto-resolve outside autopilot; PRs you don't own are untouched. Skill homes: SKILL.md (Purpose, Modes table, the "Resolve review threads" autonomy-table row, Cross-tier invariants) + reference/loop.md (safe-tier loop + its NEVER-list) + reference/feedback.md. The resolver (source-control-babysit-resolve-thread / scripts/babysit_resolve_thread.py) already exists and is only tier-gated, so this is a policy/doc change, not new tooling. It must eliminate the current SKILL-says-never-resolves vs safe-loop-says-never-skip-resolving-bot-thread contradiction in one direction.
Implementation-ordering note (does NOT gate status: ready): resolution + replies are outward GitHub writes; outward_write_mode from #440 has 0 occurrences in the skill today. Bullet 1's resolve behavior must land gated behind that mode (report-only → emit "would resolve", no mutation), so sequence bullet 1 after — or jointly with — #440.
Acceptance:SKILL.md and the safe-tier loop agree on exactly which threads the default tier resolves on your own PRs; each addressed thread ends resolved (verified via per-thread JSON action, not exit code) or carries an explicit defer + follow-up issue #; human threads never auto-resolved outside autopilot; behavior suppressed to "would resolve" under #440 report-only.
2 — Never re-trigger a review for an unchanged head SHA — already covered; no change.scripts/request_review.py (per reference/review-trigger.md) is strictly one-shot-per-head: it refuses a post when any durable request/attempt record exists for the current head SHA; a new observation window opens only on a genuinely new SHA. Dormant until the review-trigger config slots are set.
Acceptance: no behavioral change; optionally surface the invariant in the cross-tier list for visibility.
3 — Fix-round cap per head → decide-and-defer + tracked follow-up — decided (advisory-only); reconcile in #375. Operator ACCEPTED: the 3-per-head cap bounds advisory / non-convergence rounds only; blocking defects (failing CI, P0/P1, regressions) stay uncapped, preserving the existing "blocking never capped" invariant in reference/feedback.md / the Fix-Round Cap section of reference/orchestration.md. The three cap concepts (existing "max 3 CI iterations", the deliberately-high advisory ledger backstop, the new 3-per-head) are reconciled in #375, not here. On cap → decide-and-defer that files a tracked follow-up issue and posts its number (shared mechanism with bullet 1's defer path — specify once, reuse).
Acceptance: one documented cap keyed to head SHA with explicit advisory-only scope; on cap, a decide-and-defer that files a follow-up issue and reports the number; no contradiction with "blocking never capped".
4 — Merge-gate mode as a documented tier + MERGE-READY: [list] — defer to #476. Partially started: PR #456 (commit b35dd28) already landed the merge-step hard STOP. Remaining = the per-cycle MERGE-READY: [list] terminal-output convention + promoting gate-mode from prompt convention to a documented first-class tier. Planned under the merge-tier issue; not re-planned here.
5 — Parallel fan-out self-observation timing ("file after any ONE worker returns") — absorb (mechanical). Add an incremental reporting-cadence rule to the orchestration contract (reference/orchestration.md): the main agent surfaces each material observation as soon as any single worker returns, rather than withholding the report until the whole batch integrates. Strictly reporting timing — it must not alter batch dispatch/concurrency ordering (batch → integrate → next batch stays intact). Resolves the rule-9 ambiguity found in the self-audit.
Acceptance: orchestration states per-worker material outcomes are reported on first return within a cycle; concurrency-cap batching semantics unchanged.
6 — Loop-pacing keyed to worker-concurrency cap — defer to #472 (pacing/overlap). When #472 lands, its guidance's skill-home is the cadence contract (reference/cadence.md, where recommended_cadence is derived / the fixed-15m Normal ladder lives) keyed to the concurrency cap. No separate absorption planned here.
Routing
No open design call remains: bullets 1 and 3 are decided by the operator acceptance above; 2 is covered; 5 is mechanical; 4 and 6 are deferred to their own issues (#476, #472) and do not block this issue. → status: ready, priority: medium (priority: high and priority: needs-triage removed).
Label note: leaving wayfind: design in place — the operator's last curation on this issue deliberately kept it alongside the re-intake, so it reads as a durable domain tag, not an open-decision gate. Flagging rather than stripping.
Follow-up carried from #495 / PR #730 review: when the babysit skill absorbs its loop-prompt operating rules, wire the work-items permission preflight (plugins/work-items/skills/work/scripts/preflight.sh, invocation contract in plugins/work-items/reference/permission-preflight.md) as the babysit loop-start check too — the preflight ships with a documented cross-lane contract but only the work skill invokes it today; the source-control skill is under active parallel modification (#665/#675), so the wiring deliberately rides its own lane rather than a cross-plugin edit from #730.
Activity
kyle-sexton commentedon Jul 19, 2026
This was generated by AI during triage. · Category: behavioral/operational · Priority:
priority: high(week directive — pipeline improving itself). Routedneeds-human+wayfind: design.Plan: absorb v4 loop-prompt operating rules into
babysit-prsTriage of the six rules against the current skill. Two are clean (one already fully enforced, one a mechanical addition); one shared "defer → follow-up issue" mechanism is planned once and reused; two defer to existing issues; two (bullets 1 and 3) are genuinely open design and gate this plan on a maintainer decision.
A cross-cutting mechanism used by bullets 1 and 3 — on any decide-and-defer, file a tracked follow-up issue and post its number in the classification-table reply — is specified once here and referenced by both; do not plan it twice.
Per-bullet disposition
1 — Review-response ownership (fetch → fix → classification reply → RESOLVE addressed threads, or defer + follow-up issue #) — needs human. Not a simple gap: the skill contradicts itself. The authoritative tier policy states four times that the safe/default tier never resolves threads (description, purpose, per-action autonomy table, cross-tier invariants), yet the safe-tier loop's own NEVER-do list mandates "never skip resolving a BOT-authored thread" (author-conditional, after fix+reply). The classification-table reply itself is already fully specified; thread resolution via
source-control-babysit-resolve-threadexists but is gated toworker(pre-push-outdated bot threads) andautopilot(--include-human). The rule wants addressed bot/AI-review threads on your own pipeline/lane-authored PRs closed by the default invocation. Whichever direction the maintainer picks, one of the two contradicting statements must change — this is a product-intent call about the default tier's safety posture, not a mechanical absorption.resolved(verified via the wrapper's per-thread JSONaction, not exit code) or carries an explicit defer + follow-up issue #; human threads still never auto-resolved outside autopilot.2 — Never re-trigger a review for an unchanged head SHA — already covered. The review-trigger interface (
request_review.py) enforces one-shot-per-head: it refuses a post when a durable request/attempt record exists for the current head SHA, and a new observation window opens only on a genuinely new SHA. Dormant (and thus vacuously safe) until the review-trigger config slots are set. No change beyond optionally surfacing the invariant in the cross-tier list for visibility.3 — Fix-round cap per head (3) → decide-and-defer + tracked follow-up issue — needs human. Three cap concepts must be reconciled: (a) the existing "max 3 CI fix iterations per pass", (b) the advisory backstop (~high default, ledger-enforced via
manage_feedback_ledger.py record-advisory-round, deliberately not a normal limit), and (c) the proposed 3-per-head cap. The decision that makes this open: does the 3-per-head cap apply to blocking defects — reversing the deliberate, repeated invariant that blocking defects (failing CI, P0/P1, regressions) are never capped — or advisory only? No safe default exists; the two readings encode opposite convergence philosophies. Overlaps babysit-prs: codify the decide-and-defer non-convergence resolution playbook (own-once, no duplication of safety.md) #375 (non-convergence) — reconcile there, do not double-plan. The defer outcome reuses the shared follow-up-issue mechanism above.4 — Merge-gate mode as a documented tier (hard STOP +
MERGE-READY: [list]terminal output) — defer to feat(source-control): babysit autopilot merge tier — bot-account review + criteria-gated merge #476. New tier surface; theMERGE-READY: [list]per-cycle terminal-output convention and hard STOP do not exist today (only thesource-control-babysit-mergewrapper and the worker/autopilot tiers). Planned under the merge-tier issue; not re-planned here.5 — Parallel fan-out self-observation timing ("file after any ONE worker returns") — absorb (mechanical). Add an incremental reporting-cadence rule to the orchestration contract: the main agent surfaces each material observation as soon as any single worker returns, rather than withholding the report until the whole batch integrates. Scoped strictly to reporting timing — it must not alter batch dispatch/concurrency ordering (batch → integrate → next batch stays intact).
6 — Loop-pacing keyed to worker-concurrency cap (fixed 15m mismatches 30–50 min fan-out cycles) — defer to source-control:babysit-prs + loop: /loop 15m fires overlap in-flight fan-out cycles when batch-cap-10 dispatch runs long #472 (pacing/overlap). The overlap/pacing concern is squarely source-control:babysit-prs + loop: /loop 15m fires overlap in-flight fan-out cycles when batch-cap-10 dispatch runs long #472's. At most, when source-control:babysit-prs + loop: /loop 15m fires overlap in-flight fan-out cycles when batch-cap-10 dispatch runs long #472 lands, its guidance's skill-home is the cadence contract (where
recommended_cadenceis derived) keyed to the concurrency cap; no separate absorption planned here.Needs human because: bullets 1 and 3 are open design with no safe default. Bullet 1 forces a resolution of an existing SKILL.md-vs-safe-loop contradiction and decides whether the default tier resolves threads (a headline safety-posture change). Bullet 3 requires deciding whether a 3-per-head cap overrides the deliberate "blocking defects are never capped" invariant. Both are product-intent calls a maintainer must make before implementation; bullets 2/5 (already-covered / mechanical) and 4/6 (defer to #476 / #472) can proceed independently once 1 and 3 are decided.
kyle-sexton commentedon Jul 19, 2026
Elevation (2026-07-19 merge pass): thread resolution is not just hygiene — the branch ruleset has
required_review_thread_resolution, so unresolved (even fully-addressed) threads are MERGE-BLOCKING. Six digest-cleared PRs refused to merge until 9 addressed threads were manually resolved. The resolve-after-addressing checklist item in this umbrella is pipeline-critical; until it lands in the skill, the v5 lane rules carry it.kyle-sexton commentedon Jul 19, 2026
This was generated by AI (needs-human decision-brief pass).
Decision brief (2-minute read)
Problem. The triage plan cleared four of the six loop-prompt rules (two done/mechanical, bullets 4 and 6 deferred to #476/#472) but left bullets 1 and 3 open with no recommendation. This brief recommends a resolution for each; neither is settled elsewhere.
Today. The rules already run in the lane prompt, not the skill:
.work/leave-week-priorities/loop-prompts/terminal-1-babysit.md(v7) rule 3 (fetch → fix → classification reply → RESOLVE each addressed thread, or defer + follow-up issue #), rule 6b (never re-trigger an unchanged head SHA; cap fix-rounds at 3/head → decide-and-defer #375), rule 6c (MERGE-READY line), rule 9 (file after any one worker returns). Skill home:plugins/source-control/skills/babysit-prs/. The contradiction the triage names is real — SKILL says the safe/default tier never resolves threads, while the safe-tier loop's NEVER-list mandates "never skip resolving a BOT-authored thread"; the resolver exists (bin/source-control-babysit-resolve-thread,scripts/babysit_resolve_thread.py) but is gated to worker/autopilot.Decisive fact (already in this issue's elevation comment — build on it, not restated). The branch ruleset carries
required_review_thread_resolution, so addressed-but-unresolved threads are merge-blocking; six digest-cleared PRs refused to merge until threads were resolved by hand. That breaks the bullet-1 tie.Bullet 1 — does the default tier resolve threads?
Bullet 3 — does the 3-per-head cap apply to blocking defects?
Cross-links. #440 (
outward_write_mode: report-onlydefault) directly constrains bullet 1: thread resolution and classification replies are outward GitHub writes, so the resolving behavior must be gated behindwritemode and suppressed under report-only (report "would resolve" without mutating). #476 (merge-tier — bullet 4's MERGE-READY gate-mode deferred there; APPROVED staged #450 → #373/#327/#326 → implement). #472 (pacing — bullet 6). #375 (non-convergence — bullet 3 reconciliation home). #450 (bot identity). Sibling absorption umbrellas #478/#479.kyle-sexton commentedon Jul 19, 2026
Operator decision (2026-07-19, decision-session batch 2; relayed).
ACCEPTED — the decision brief's RECOMMENDED option verbatim.
Re-intake: removing
needs-human, addingpriority: needs-triagefor routing.kyle-sexton commentedon Jul 19, 2026
This was generated by AI during triage.
T2 routing plan — decided-state implementation map
Re-intake pass after the operator acceptance (
2026-07-19T22:35, decision-session batch 2): the RECOMMENDED options in the decision brief were accepted verbatim, which closes the only two open design calls (bullets 1 and 3). This comment does not re-derive that rationale — see the decision brief above — it consolidates the now-decided state into an implementation map and routes.North-star note (sort, not tier): this is a pipeline self-improvement item and is relevant to the week's throughput north-star, so it should sort early within its tier — but per the
#552rebaseline that alone does not earnpriority: high(not blocking other work, not scaling-chain). Routedpriority: medium.Per-bullet disposition (skill home · behavioral contract · acceptance)
1 — Review-response ownership → RESOLVE addressed threads — decided (absorb + fix contradiction). Operator ACCEPTED: on pipeline/lane-authored PRs the default tier, after fix + classification-table reply, resolves each addressed bot/AI-review thread it owns; human threads still never auto-resolve outside autopilot; PRs you don't own are untouched. Skill homes:
SKILL.md(Purpose, Modes table, the "Resolve review threads" autonomy-table row, Cross-tier invariants) +reference/loop.md(safe-tier loop + its NEVER-list) +reference/feedback.md. The resolver (source-control-babysit-resolve-thread/scripts/babysit_resolve_thread.py) already exists and is only tier-gated, so this is a policy/doc change, not new tooling. It must eliminate the current SKILL-says-never-resolves vs safe-loop-says-never-skip-resolving-bot-thread contradiction in one direction.status: ready): resolution + replies are outward GitHub writes;outward_write_modefrom#440has 0 occurrences in the skill today. Bullet 1's resolve behavior must land gated behind that mode (report-only → emit "would resolve", no mutation), so sequence bullet 1 after — or jointly with —#440.SKILL.mdand the safe-tier loop agree on exactly which threads the default tier resolves on your own PRs; each addressed thread endsresolved(verified via per-thread JSONaction, not exit code) or carries an explicit defer + follow-up issue #; human threads never auto-resolved outside autopilot; behavior suppressed to "would resolve" under#440report-only.2 — Never re-trigger a review for an unchanged head SHA — already covered; no change.
scripts/request_review.py(perreference/review-trigger.md) is strictly one-shot-per-head: it refuses a post when any durable request/attempt record exists for the current head SHA; a new observation window opens only on a genuinely new SHA. Dormant until the review-trigger config slots are set.3 — Fix-round cap per head → decide-and-defer + tracked follow-up — decided (advisory-only); reconcile in
#375. Operator ACCEPTED: the 3-per-head cap bounds advisory / non-convergence rounds only; blocking defects (failing CI, P0/P1, regressions) stay uncapped, preserving the existing "blocking never capped" invariant inreference/feedback.md/ the Fix-Round Cap section ofreference/orchestration.md. The three cap concepts (existing "max 3 CI iterations", the deliberately-high advisory ledger backstop, the new 3-per-head) are reconciled in#375, not here. On cap → decide-and-defer that files a tracked follow-up issue and posts its number (shared mechanism with bullet 1's defer path — specify once, reuse).4 — Merge-gate mode as a documented tier +
MERGE-READY: [list]— defer to#476. Partially started: PR#456(commitb35dd28) already landed the merge-step hard STOP. Remaining = the per-cycleMERGE-READY: [list]terminal-output convention + promoting gate-mode from prompt convention to a documented first-class tier. Planned under the merge-tier issue; not re-planned here.5 — Parallel fan-out self-observation timing ("file after any ONE worker returns") — absorb (mechanical). Add an incremental reporting-cadence rule to the orchestration contract (
reference/orchestration.md): the main agent surfaces each material observation as soon as any single worker returns, rather than withholding the report until the whole batch integrates. Strictly reporting timing — it must not alter batch dispatch/concurrency ordering (batch → integrate → next batch stays intact). Resolves the rule-9 ambiguity found in the self-audit.6 — Loop-pacing keyed to worker-concurrency cap — defer to
#472(pacing/overlap). When#472lands, its guidance's skill-home is the cadence contract (reference/cadence.md, whererecommended_cadenceis derived / the fixed-15m Normal ladder lives) keyed to the concurrency cap. No separate absorption planned here.Routing
No open design call remains: bullets 1 and 3 are decided by the operator acceptance above; 2 is covered; 5 is mechanical; 4 and 6 are deferred to their own issues (
#476,#472) and do not block this issue. →status: ready,priority: medium(priority: highandpriority: needs-triageremoved).Label note: leaving
wayfind: designin place — the operator's last curation on this issue deliberately kept it alongside the re-intake, so it reads as a durable domain tag, not an open-decision gate. Flagging rather than stripping.feat(ci): add skill-quality static gate lane for changed skills (#530) (
fix(source-control): map babysit recommended_cadence to ScheduleWakeu…
kyle-sexton commentedon Jul 20, 2026
Follow-up carried from #495 / PR #730 review: when the babysit skill absorbs its loop-prompt operating rules, wire the work-items permission preflight (plugins/work-items/skills/work/scripts/preflight.sh, invocation contract in plugins/work-items/reference/permission-preflight.md) as the babysit loop-start check too — the preflight ships with a documented cross-lane contract but only the work skill invokes it today; the source-control skill is under active parallel modification (#665/#675), so the wiring deliberately rides its own lane rather than a cross-plugin edit from #730.
feat(work-items): extract shared self-observation filing contract (#540…
fix(work-items): re-key triage scope on state, not authorship (#486) (#…
feat(claude-ops): script deterministic lane-cycle mechanics (#538) (#803
kyle-sexton commentedon Aug 13, 2026
Frontier disposition (drain session)
Ruling: Close as completed.
v4 loop-prompt operating rules absorption is tracked in source-control/babysit-loop skill conventions.