Compress post-review-pickup liveness workflow

Context

This ticket exists because recent skill-work tried to repair valid Agent OS liveness friction by adding more prose to already-hot skill payloads.

Live evidence from this session:

• PR #13530 was closed unmerged after the operator challenged its additive skill-work shape. Its diff added 30 lines and removed 1 line across .agents/skills/post-review-pickup/** and .agents/skills/ticket-intake/**.
• node ./ai/scripts/lint/lint-skill-manifest.mjs --report-sizes --top 15 reports files=103 bytes=617311 lines=7353 for skill Markdown.
• The hot workflow payloads are already large:
  • .agents/skills/pr-review/references/pr-review-guide.md = 50,230 bytes
  • .agents/skills/pull-request/references/pull-request-workflow.md = 34,096 bytes
  • .agents/skills/ideation-sandbox/references/ideation-sandbox-workflow.md = 25,000 bytes
  • .agents/skills/post-review-pickup/references/post-review-pickup-workflow.md = 23,131 bytes

The smallest-context active peer still pays this substrate cost most sharply. Adding more "BOOKS" makes GPT/Codex lifecycle work less reliable, not more reliable.

Release classification: post-release hardening / Agent OS stability; boardless.

The Problem

The liveness problems behind #13523, #13529, and #13530 are real: agents should not silently idle out, duplicate stale work, or pile more author PRs onto a scarce review bottleneck. The failed shape is encoding each symptom as another local paragraph inside hot workflow payloads.

That creates a recurring negative loop:

1. A workflow miss happens.
2. We add another instruction paragraph.
3. The workflow becomes heavier to load.
4. Smaller-context agents lose more working room.
5. Lifecycle completion gets harder, so more workflow misses happen.

This ticket is for the inverse move: compress the hot post-review-pickup liveness workflow and preserve the contracts through clearer consolidation, not accretion.

The Architectural Reality

The relevant substrate is .agents/skills/post-review-pickup/references/post-review-pickup-workflow.md.

That file is a hot lifecycle payload. It repeats related concepts across heartbeat handling, reviewer handoff, lane-state:, lifecycle boundaries, authoring lanes, externally falsifiable terminals, broadcast-suppressed fallback, and examples.

The create-skill guide already states the governing rule: workflow files are maps, and edge-case sections should compress to one-line triggers or move only when the total substrate actually gets smaller. It also says to measure bloat against the smallest-context peer.

The Fix

Compress the post-review-pickup workflow by merging duplicated terminal/liveness wording and collapsing symptom-specific additions into existing general routing rules.

The implementation should:

• Keep SKILL.md unchanged unless a one-line trigger correction is strictly required.
• Prefer editing the existing workflow over adding new files.
• Avoid "net-reduce by relocation" where the same text just moves into a sibling payload.
• Preserve the actual liveness contracts from recent work:
  • Heartbeat/watchdog wakes run the lifecycle cycle.
  • Every lifecycle boundary emits an explicit lane-state:.
  • Valid terminals remain externally falsifiable.
  • Awareness wakes do not manufacture duplicate work when live state is already handled.
  • Reviewer scarcity routes toward coordination instead of reflexively adding another implementation PR.
  • Broadcast suppression is not a terminal.

Contract Ledger Matrix

Decision Record impact

Aligned with ADR 0008 (SKILL.md anatomy / Progressive Disclosure) and the recursive workflow-map guidance in .agents/skills/create-skill/references/skill-authoring-guide.md.

Acceptance Criteria

• Record before/after byte counts for the touched skill Markdown files.
• The PR has a net-negative .agents/skills/**/*.md byte delta.
• .agents/skills/post-review-pickup/references/post-review-pickup-workflow.md is reduced by at least the additive footprint PR #13530 attempted to add, unless the PR body documents a stronger measured reduction elsewhere in the same skill payload.
• No new skill Markdown file is added unless the aggregate skill Markdown delta remains negative and the PR explains why the new file is a true progressive-disclosure boundary rather than relocation.
• Existing liveness contracts listed in The Fix remain present and grep-verifiable.
• node ./ai/scripts/lint/lint-skill-manifest.mjs --base origin/dev passes.

Out of Scope

• Reopening or reworking PR #13530.
• Changing the lifecycle policy itself.
• Adding a new skill or another workflow book.
• Compressing pr-review, pull-request, or ideation-sandbox; those are separate candidates after this first focused reduction.

Avoided Traps

• Symptom paragraph per incident: rejected because it is the bloat loop this ticket exists to break.
• Relocation-only extraction: rejected because smaller-context agents still pay the same aggregate substrate cost once the workflow is invoked.
• Weakening liveness to save bytes: rejected; the point is consolidation, not behavioral retreat.

Related

• Parent: #10757
• Guard: #13533
• Closed additive precedent: #13530
• Related closed additive/drop lessons: #13523, #13529

Handoff Retrieval Hints

Search patterns:

• skill bloat workflow markdown growth PR 13530 post-review-pickup
• smallest-context peer 258k skill payload bloat
• post-review-pickup externally falsifiable terminal lane-state heartbeat