Author the blog-post authoring skill — narrative arc, source-every-claim, no-overclaim, cross-family review bar

Context

Sub of #13383 (the v13 hero-piece blog stream). The stream will produce many posts, but there is no authoring skill — so each post re-derives the discipline from scratch. Empirically: #13486 (the cross-family-verification post) took multiple cross-family review cycles to converge, catching three distinct over-claim flavors. This ticket codifies that discipline so the next post starts with it instead of earning it through review.

Dup-sweep (2026-06-21): searched open+closed for "blog post skill" / "blog authoring narrative" + listed .agents/skills/ — no blog-authoring skill exists. Closest: #13383 (the content epic — this ticket's parent), #10321 (release-WORKFLOW skill = release notes + publish, distinct), update-roadmap (sibling beat). The release-notes hero-piece methodology currently lives only in agent memory; this skill is its blog sibling.

The Problem

Blog-post authoring has a high, repeatable failure surface #13486 surfaced first-hand:

• weak narrative arc (changelog-dump vs hero-piece);
• unsourced external claims (a verify-before-assert failure — RC'd twice on #13486);
• over-claims in three distinct flavors (unsourced superlative; universal quantifier "all N are X"; misleading fraction/framing "129/151 ⇒ almost done") — literally-true-but-imply-false;
• skipped/insufficient cross-family review (the structural catch for all of the above). Without a skill, every post re-incurs these.

The Architectural Reality

A Progressive-Disclosure skill at .agents/skills/blog-post/: a 7–12-line SKILL.md router (trigger = authoring/revising a public blog post) + references/blog-authoring-guide.md payload. Edge-case-triggered (loaded only when a post is authored) → low per-turn cost. Sibling of the release-notes methodology + update-roadmap.

The Fix

Author the skill, distilling #13486's first-hand context:

1. Narrative arc — hero-piece structure (TL;DR thesis → hook → arc → receipts/proof → CTA); lead with the thesis, not the volume hook; release-notes level.
2. Source every external claim — verify-before-assert; an authority's/operator's verbal statement is NOT a citable source; cut or independently verify before publish.
3. The three over-claim flavors — unsourced superlative, universal quantifier, misleading fraction/framing; claims must match reality in implication, not just literal accuracy.
4. Cross-family review bar — 2+ model reviews mandatory; the authority/operator approves LAST (so peers don't anchor and rubber-stamp); the review is the structural backstop, not a formality.

Acceptance Criteria

• .agents/skills/blog-post/SKILL.md router (7–12 lines, YAML name/description, trigger language) + references/blog-authoring-guide.md payload.
• Payload covers the four disciplines, each anchored to the #13486 empirical case.
• skills.manifest.json updated (name/description mirror + budgets) + .claude/skills/blog-post symlink.
• node ai/scripts/lint/lint-skill-manifest.mjs --base origin/dev passes.
• The two skill-PR-open gates: Contract Ledger on this ticket + /turn-memory-pre-flight load-effect audit in the PR body.
• Cross-family review (a skill about cross-family review, reviewed cross-family — fitting).

Out of Scope

• The release-WORKFLOW skill (#10321 — release notes + publish).
• Producing specific posts (#13383's stream; this is the enabling substrate).

Decision Record impact

none — new workflow skill following ADR 0008 (skill anatomy); no new ADR.

Related

Parent: #13383. Sibling: update-roadmap + the release-notes methodology. Empirical anchor: #13486 / #13485 (the cross-family-verification blog — authored + cross-reviewed this session; every over-claim flavor caught + fixed). Authoring contract: ADR 0008 / create-skill.

Origin Session ID: c4fcedd0-c449-4f8c-b368-e3ac0c0509ff Retrieval Hint: "blog-post authoring skill narrative arc source-every-claim no-overclaim three over-claim flavors cross-family review bar #13383 stream #13486 anchor"