agentleFS
Sign inStart free
powerczukpowerczuk/seo-machine/skills/citable-passage/SKILL.md

citable-passage

Skill0 starsChanged 2 months agoMITView on GitHub

Score passage-level extractability for AI Overviews / AI Mode (density, heading alignment, definitions, attribution, sub-query coverage) and return rewrites.

The file

Create a free account to mount
Files
SKILL.md

/citable-passage — Passage extractability scorer

Target: $ARGUMENTS

Deep dive behind criterion G6. Evidence: passage scoring (US9940367B1, US9959315B1) and generative summaries (US11769017B1) — ${CLAUDE_PLUGIN_ROOT}/skills/seo-fundamentals/references/patent-signals.md. Generative search decomposes a query into sub-queries, retrieves passages per sub-query, and cites passages that are self-contained, factual, and attributable. An 8,000-word article of rambling prose loses to a 1,500-word article structured as answers.

Inputs

Article (URL → python3 ${CLAUDE_PLUGIN_ROOT}/scripts/fetch_page.py "<url>" --json; file or paste → save and run text_metrics.py for word count and the passage-candidate count), primary query (infer from title/H1 if absent and say so).

Rubric (100)

ModulePtsHow to score
1. Passage density20Self-contained passages per 1,000 words: 1–3 sentences (≤60 words), complete claim, supporting fact (number/entity/date), meaningful in isolation (no "as mentioned above"). ≥6 → 20 · 4–5 → 15 · 2–3 → 10 · 1 → 5 · 0 → 0. Start from the script's candidate count, then verify by reading.
2. Heading-to-content alignment20Share of H2/H3 phrased as a question or extractable answer whose first 40–60 words answer it. (aligned ÷ total) × 20.
3. Definition-style openings15Share of concept sections opening with "X is Y that does Z". ≥80% → 15 · 60–79 → 10 · 40–59 → 5 · <40 → 0.
4. Attribution15Share of major claims with inline, specific, primary-source attribution. (attributed ÷ claims) × 15.
5. Sub-query coverage30Generate 8–12 fan-out sub-queries (definition, how-to, comparison, troubleshooting, best practice, alternatives, cost, when). (answered extractably ÷ total) × 30.

Verdict (from scoring-spec.md): EXTRACTABLE 80–100 · PARTIAL 60–79 · WEAK 40–59 · OPAQUE 0–39. Maps to G6: EXTRACTABLE → 2, PARTIAL → 1, else 0. The article-level "AI Overviews readiness" verdict is computed by /score from Category G, not here.

Fixes

For each weak module produce concrete rewrites: split long paragraphs into claim + fact; move the answer to the first sentence after the heading; rewrite concept openings as definitions; replace vague attributions with named primary sources (only sources that exist — never invent a citation); add or expand sections for uncovered sub-queries. Keep every real fact; add none.

Output

  1. Scorecard (5 modules + total) and verdict.
  2. Sub-query coverage table (sub-query · covered · extractable · well-answered).
  3. Top 10 fixes ordered by impact, each with the rewritten passage.
  4. Re-test: run /citable-passage again after edits.

Next: /score for the full article verdict, /alignment-audit before deploying schema, humanizer for the final pass.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free