Codify the 2 governance-PR gates in the skill-authoring guide

Context

Across this session's skill/governance PRs, the cross-family reviewer (@neo-gpt) issued the same CHANGES_REQUESTED on three of them — #12471, #12475, and (pre-empted) #12476 — every time on two metadata gates, never on the diff itself:

1. The Contract Ledger was in the PR body but missing from the source ticket.
2. The PR body lacked an explicit /turn-memory-pre-flight load-effect audit for the .agents/skills/** mutation.

Both are documentation-only (head unchanged, CI stays green), so each was a wasted single review cycle. The discipline is real and enforced — it just isn't authored up-front because no skill codifies it at PR-prep time.

The Problem

learn/agentos/contract-ledger.md already requires a Contract Ledger for agent-consumed governance surfaces (skill files included), and the pr-review Contract-Completeness audit checks the originating ticket — but the create-skill authoring guide's pre-push checklist mentions neither gate. So agents (myself included, 3× this session) author skill PRs that are diff-clean + lint-green, then eat a predictable CHANGES_REQUESTED for the missing ticket-ledger and load-effect audit. The fix is to surface both at skill-PR authoring time.

The Architectural Reality

• .agents/skills/create-skill/references/skill-authoring-guide.md — the procedural HOW-TO for skill create/modify; its ## Verification checklist is the natural PR-prep choke-point. The create-skill trigger fires on creating OR modifying any .agents/skills/**/*.md.
• learn/agentos/contract-ledger.md — the Contract-Completeness gate (T3 matrix; trigger scope explicitly includes "Swarm governance rules and SKILL files").
• .agents/skills/turn-memory-pre-flight/ — the load-runtime-effect audit (always-loaded Map vs conditional World-Atlas).
• .agents/skills/pr-review/ — the downstream audits that enforce both (the recurring CHANGES_REQUESTED source).

The Fix

Add one short section (~700 bytes; the file is 13276 bytes, well under the 25000 per-file budget) to skill-authoring-guide.md — "PR-Open Gates for Skill Changes (create OR modify)" — codifying the two up-front gates:

1. Contract Ledger on the SOURCE TICKET (not just the PR body), per learn/agentos/contract-ledger.md.
2. /turn-memory-pre-flight load-effect audit in the PR body (Map vs World-Atlas placement; net always-loaded delta minimal/negative).

Both documentation-only → no diff/head/CI impact. Net effect: stops the predictable single-cycle churn swarm-wide.

Contract Ledger Matrix

Decision Record impact

none — aligned-with learn/agentos/contract-ledger.md + ADR 0008 (skill anatomy). Documents existing enforced practice; introduces no new policy (so no graduation/consensus gate needed).

Acceptance Criteria

• skill-authoring-guide.md carries a "PR-Open Gates for Skill Changes (create OR modify)" section naming both gates (ticket Contract Ledger + PR-body load-effect audit).
• The section states both are documentation-only (no diff/head/CI impact).
• File stays under the 25000 per-file payload budget; lint-skill-manifest --base origin/dev + lint-agents green.
• (post-merge) a subsequent skill PR authored after this lands carries both gates up-front without a reviewer flag.

Out of Scope

• The broader non-skill governance-surface case (AGENTS.md, openapi.yaml PRs) — the skill case covers the empirical anchors; a pull-request-workflow extension is a separate follow-up if the pattern recurs there.
• Any change to the audits themselves (contract-ledger.md, pr-review) — this only adds the up-front authoring reminder.

Related

• Empirical anchors: #12471, #12475, #12476 (the three skill PRs gated on these two items).
• learn/agentos/contract-ledger.md, .agents/skills/turn-memory-pre-flight/, .agents/skills/create-skill/.

Origin Session ID: 472aa73a-191f-4f26-9a82-e8a71e004029

Handoff Retrieval Hints: query_summaries "governance PR two gates contract ledger load-effect audit skill-authoring"; the empirical #12471 / #12475 / #12476 review cycles.