Wire agent-preflight discovery and PR-body anchor parity

Context

PR #13842 shipped the standalone agent-preflight helper for #13813 and received an Approve+Follow-Up review. The helper composes the existing pre-commit / PR-body checks correctly, but the review surfaced two non-blocking residuals that should not stay buried in the PR thread:

1. Agents are not yet routed to discover and run agent-preflight, so #13813's "cut commit-retry rounds" value remains latent.
2. The helper mirrors PR-body anchor lists from .github/workflows/agent-pr-body-lint.yml, but that sync-by-convention surface needs an explicit parity guard or source pointer.

Release classification: post-release / boardless (Approve+Follow-Up follow-up — non-blocking for #13842).

Duplicate sweep: checked the latest 30 open GitHub issues at 2026-06-22T02:20Z and found no equivalent follow-up. A2A in-flight sweep checked the recent 50-message slice and found no competing [lane-claim] / [lane-intent] for this scope. Exact local sweep for agent-preflight, agent-pr-body-lint, prReviewAnchors, and commit-retry terms found #13813 plus older #9864 / #11501 / #11502 anchor-history context, but no ticket for wiring the newly shipped helper into agent workflow discovery.

The Problem

#13813 solves the repeated commit-retry loop only if agents actually know to run the helper before committing or opening a PR. A tool discoverable only through package.json can pass tests and still miss its MX outcome.

There is also one accepted duplication surface: PR-body anchors are intentionally mirrored by convention because the prior shared-module extraction (prReviewAnchors.mjs) was reverted in #11502 after #11501. Re-proposing extraction is the wrong shape; the follow-up should strengthen the accepted sync-by-convention shape instead.

The Architectural Reality

• agent-preflight is opt-in developer tooling, not a husky or CI gate extension.
• The workflow authority for PR body anchors remains .github/workflows/agent-pr-body-lint.yml; shared extraction is explicitly rejected by the workflow comment history.
• Agent routing lives in the agent workflow / skill substrate, so discovery should be added where commit / PR pre-flight behavior is already described, without bloating turn-loaded surfaces more than necessary.

The Fix

Add the smallest wiring that makes the helper discoverable and keeps the mirrored PR-body anchors honest:

• Add a concise pointer from the appropriate commit / PR pre-flight guidance to run agent-preflight before the first commit / PR body attempt.
• Add either a source-pointer comment or a focused parity test that ties the helper's PR-body anchor lists to .github/workflows/agent-pr-body-lint.yml as the source-to-mirror.
• Do not reintroduce a shared prReviewAnchors.mjs module or rewrite the live workflow gate.

Contract Ledger Matrix

Decision Record impact

None. This is workflow discoverability and drift-guarding for a shipped build helper; it does not amend ADR authority.

Acceptance Criteria

• Agent workflow / pre-flight guidance points agents to run agent-preflight before the first commit / PR-body attempt when working on agent-authored PRs.
• The helper's PR-body anchor mirror has either a focused parity test or an explicit source-pointer comment to .github/workflows/agent-pr-body-lint.yml.
• The implementation does not reintroduce a shared prReviewAnchors.mjs module and does not rewrite the live GitHub Action gate.
• Focused validation for the touched helper/docs/tests passes.

Out of Scope

• Changing husky hooks or CI gate behavior.
• Reopening the reverted shared-anchor extraction shape from #11501 / #11502.
• Expanding #13842 before human merge; this is a follow-up lane after that PR lands or a stacked follow-up if the operator explicitly wants it.

Related

• #13813
• PR #13842
• #9864
• #11501
• #11502

Origin Session ID: 43e8ab91-e980-4303-b70f-898d8c4ae98e

Retrieval Hint: agent-preflight discovery wiring PR-body anchor parity #13813 #13842 prReviewAnchors #11501 #11502