Conciseness

Dense rule-per-line style with zero explanation of concepts Claude already knows and no padding. Not 5 because several hard-nos repeat verbatim across sections ('Do not compose autopilot-pr-merge-worker' appears four times; 'never comments, labels, assigns, or requests reviewers' three times; 'Do not list all open PRs' twice) — deliberate fail-closed emphasis, but trimmable.

Actionability

Two complete executable gh pr list commands with exact JSON fields, plus exact activation-card, exit-receipt, and queue-table templates with literal rationale strings. Not 5 because key mechanics lack commands: snapshotting the CODEOWNERS set from reviewRequests, paginating comments/reviews to exhaustion, and probing the sibling worker on disk are described but not executable as written.

Workflow Clarity

Clear sequenced Procedure, an 8-step CODEOWNERS gate with explicit fail-closed branches and drop rationales, and the mandatory queue table acting as a validation checklist ('Missing table, missing column, or blank rationale -> stop'), plus empty-keep-set and stop-condition handling. This is a batch fan-out operation and validation checkpoints are thoroughly present, so the batch cap does not apply.

Progressive Disclosure

The one bundle reference, assets/fan-out-pool.md, is real, clearly signaled, one level deep, and appropriately holds the pool detail; section headers are clean. Not 5 because the ~210-line body keeps the long CODEOWNERS gate and queue-table spec fully inline, content that could arguably live in a reference file, leaving minor organization gaps.

Specificity

Concrete actions with parameters: 'Queue microsoft/apm pull requests labelled `panel-review`', 'fan them out through an isolated pool (default 2) of autopilot-pr-review-worker sessions', plus explicit scope exclusions. It sits at anchor 4 rather than 5 because what the review itself produces (advisory panel review) is only implied via the worker skill name, leaving a minor coverage gap.

Completeness

The 'what' is clear (queue labelled PRs, fan out through an isolated worker pool, advisory only), but there is no 'Use when...' clause or equivalent explicit trigger guidance — the guideline caps completeness at 3 when the 'when' is only weakly implied (here, only via the label name). The environment list ('Works in a local session, Copilot App automation...') describes where, not when to invoke.

Trigger Term Quality

Good natural keyword coverage: 'pull requests', 'review', 'queue', 'panel-review', 'fan out', plus named execution environments. Not 5 because common variations users might say ('code review', 'review the PRs', an explicit 'PR review' task phrase) are absent; not 3 because coverage goes well beyond 'some relevant keywords'.

Distinctiveness Conflict Risk

Clear niche with distinct triggers: specific repo, specific label, specific worker skill name, and explicit disambiguation from siblings ('Never composes autopilot-pr-merge-worker', 'Does not triage issues or open greenfield PRs'). Minimal conflict risk even within the scheduler family.