# Github Skill Adoption Audit

> Audit GitHub repositories that publish agent skills and summarize real adoption signals: stars, forks, watchers, issues, recent views, clones, referrers, popular paths, directory/marketplace indexing, and external references. Use when Codex is asked whether a skill repo has users, how many stars it has, whether people are cloning it, which repo is getting traction, or how to prioritize skill promotion.

- Skill: `walterkken/github-skill-adoption-audit` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add walterkken/github-skill-adoption-audit`
- Raw SKILL.md: https://api.skillmd.com/api/skills/walterkken/github-skill-adoption-audit/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: walterkken (https://skillmd.com/u/walterkken)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/walterkken/github-skill-adoption-audit

---


# GitHub Skill Adoption Audit

## Overview

Measure whether a public skill repository is being discovered or used. Separate
hard signals such as stars and unique cloners from weak signals such as generic
search hits.

Read `references/adoption-signals.md` when interpreting results.

## Workflow

1. Resolve the repository set.
   - If the user names a repo, inspect only that repo.
   - If the user says "my skills", discover likely skill repos from local git
     remotes or `gh repo list <owner>`.

2. Gather GitHub metadata.
   - Use `gh repo view` for stars, forks, watchers, issues, PRs, visibility,
     pushed date, and URL.
   - Use GitHub traffic endpoints for views, clones, referrers, and popular
     paths when the authenticated user has permission.

3. Check external discovery.
   - Search for exact repo names, install commands, and unique skill phrases.
   - Treat generic skill names as weak evidence unless the result clearly points
     back to the user's repo.
   - Note GitHub code-search rate limits rather than retrying aggressively.

4. Interpret cautiously.
   - Stars show public endorsement.
   - Unique clones suggest installation or inspection, but may include bots,
     mirrors, CI, or the owner.
   - Views/referrers show discovery channels for the last GitHub traffic window,
     usually about 14 days.
   - Marketplace indexing is discoverability, not proof of active use.

5. Output a short decision.
   - Identify the strongest repo.
   - List star count and traffic counts.
   - Say what is not yet proven.
   - Recommend one practical next action: README demo, screenshot, release,
     example task, marketplace metadata, or targeted distribution.

## Script

Run the bundled audit script when `gh` is authenticated:

```bash
python3 scripts/audit_skill_repos.py \
  --owner walterkken \
  --repos doc-driven-engineering-software-skills matlab-simulink-agent-skill \
  --output skill-adoption-audit.md
```

It writes a markdown report and a sibling JSON file. If traffic endpoints fail,
the report should still include public metadata and the exact permission/rate
limit issue.