It isn't being found. First learn why, then rewrite the surfaces.

One install chains two skills: Findable audits the surfaces people actually search — README opening, GitHub About text, topics, site page — against the task queries a stranger would type, and Wordsmith writes two or three alternatives per finding that fix it without inventing a fact. You get the diagnosis and the cure in one pass.

2 chained skills 5-10 target queries 2-3 options per issue 0 facts invented

for coding agents · free, open, MIT

install

Copy one line. Audit the surface, then rewrite it.

The install line from the repo README. This skill is a chain: findable and wordsmith must be installed too — if either is missing it says so and stops.

npx skills add arm-pz/findable-polish --skill findable-polish

the workflow

Three phases. The audit output arrives verbatim.

Phase one runs the Findable audit and hands its full report on, untouched — no summarising the evidence away. Phase two turns each finding into alternatives. Phase three returns everything in one copy-paste-ready block.

Audit 1

Findable reads the target content and derives the queries a stranger types, task-based, never name-based.

  • 5-10 target queries
  • missing keywords
  • vague or audience-less description
  • weak opening line
  • absent About text or topics
  • same-name collisions

Alternatives 2

Wordsmith treats each audit finding as its own rewrite task, in the author's voice and facts.

  • 2-3 labelled options per issue
  • keywords woven in naturally
  • GitHub About and topic drafts
  • opening lines matched to queries
  • best option chosen with a reason

Presentation 3

One three-part response: audit summary by severity, alternatives grouped by issue, and a merge recommendation.

  • severity-ordered findings
  • copy-paste-ready blocks
  • a 2-3 sentence overall recommendation
  • unchanged facts stay untouched

Outside its boundary 2 siblings

The chain stops where its parts begin. Tuning the SKILL.md description that carries agent triggering is tune-skill. Setting up a repo's discovery from scratch, with distribution targets, is findable. This skill writes human-facing copy from a real audit; it does not execute gh commands or push changes — it hands over the commands.

  • audits + drafts, never publishes
  • no trigger tuning
  • no resume work (recruiter-lens)
  • no code review

ground rules

What it refuses to do.

Queries before copy

Every rewrite is written from a search query a real stranger would type, derived and shown before any prose. Copy written from vibes gets audited again by the search engine, not by you.

Describe only what is real

Alternatives may not claim features, install counts or compatibility the content doesn't support. A rewrite that adds a false capability is a worse audit result, not a better one.

Hand the audit over whole

Phase two receives the full phase-one report verbatim. Summarising findings before rewriting them drops the awkward ones, and the awkward ones are the point.

Both the diagnosis and the cure, in one pass.

Free, open, and every alternative traces back to a finding the audit named.

Install findable-polish Read the source