GitHub Starred Knowledge Base (pedroiff0)

This skill turns the user's 41 starred GitHub repositories into a reusable knowledge base. Instead of cloning 41 repos (huge, stale fast), it keeps a mapped index and fetches the specific repo's content on demand via the GitHub raw API / gh api when a query matches a domain.

Use it whenever the user asks for: a free API, a self-hosted tool, a Python library, an AI agent/RAG pattern, an algorithm, a math/edu resource, an astronomy reference, a Node/web example, or anything their starred repos cover.

The 41 starred repos, by domain

Full table (with stars + raw URL) lives in references/starred-index.md. Quick domain map:

How to use (on-demand retrieval)

When a query falls in a domain, fetch the relevant repo's README or specific file via raw GitHub, then synthesize. Do NOT clone the whole star list.

1. Resolve which starred repo(s) match

# list domains for a keyword
python3 skills/github/github-starred-kb/scripts/fetch_kb.py --match "api clima"
# -> public-apis/public-apis

2. Fetch the raw README (or a path)

# via helper (prints raw URL or content)
python3 skills/github/github-starred-kb/scripts/fetch_kb.py --repo public-apis/public-apis --readme
# manual equivalent:
curl -sL https://raw.githubusercontent.com/public-apis/public-apis/master/README.md

3. For "awesome" lists, grep the section you need

curl -sL https://raw.githubusercontent.com/public-apis/public-apis/master/README.md \
  | grep -iA3 "weather\|climate"

4. For the user's own repos, prefer the local clone when present

# e.g. financas-app lives at /home/pedro/Repositorios/pessoal/financas-app
# awesome-skills at /home/pedro/Repositorios/pessoal/awesome-skills

Refreshing the index

The star list changes over time. To rebuild references/starred-index.md:

python3 skills/github/github-starred-kb/scripts/fetch_kb.py --rebuild

This calls gh api users/pedroiff0/starred and rewrites the index with current stars + categories (categories are heuristics; review the diff).

Pitfalls

• Don't clone all 41. They're large and go stale; fetch on demand.
• Raw default branch differs (master vs main). The helper tries both; if a manual curl 404s, swap the branch.
• awesome lists are huge — grep the specific section, don't paste the whole file.
• Categories are a heuristic map, not gospel; a repo can serve multiple domains.
• Prefer the local clone for the user's own repos (fresher + offline).