scout-repos

Scout Repos

Review curated public repos for Claude Code configuration patterns worth adopting.

Step 0: Discover tools

WebFetch and WebSearch are deferred tools — they may not be loaded yet. Run ToolSearch("WebFetch WebSearch") to ensure both tools are available before proceeding.

Step 1: Load sources

Read $DOTFORGE_DIR/practices/sources.yml. Build a list of repos with their focus areas.

Step 2: Fetch each repo's config

For each source, use gh CLI to fetch their Claude Code configuration files:

# Fetch CLAUDE.md
gh api repos/{owner}/{repo}/contents/CLAUDE.md --jq '.content' | base64 -d 2>/dev/null

# Fetch .claude/ directory listing
gh api repos/{owner}/{repo}/contents/.claude --jq '.[].name' 2>/dev/null

# Fetch specific files
gh api repos/{owner}/{repo}/contents/.claude/settings.json --jq '.content' | base64 -d 2>/dev/null
gh api repos/{owner}/{repo}/contents/.claude/rules --jq '.[].name' 2>/dev/null

If gh fails (rate limit, private repo), try WebFetch on the raw GitHub URL:

https://raw.githubusercontent.com/{owner}/{repo}/main/CLAUDE.md
https://raw.githubusercontent.com/{owner}/{repo}/main/.claude/settings.json

If both fail, skip the repo and note it in the report.

Step 3: Analyze patterns

For each repo's config, extract patterns in the source's focus areas:

Hooks

• New hook event/matcher combinations not in dotforge
• Novel blocking logic (different from block-destructive or lint-on-save)
• Hook chaining patterns

Settings

• Deny list entries we don't have
• Permission patterns (granular vs wildcard)
• Hook wiring approaches

Rules

• Globs patterns we don't use
• Rule content structure (frontmatter fields, sections)
• Cross-file rule dependencies

Agents

• Agent definitions with novel roles
• Memory policies different from ours
• Orchestration patterns

Commands/Skills

• Custom commands we could generalize
• Skill structures with novel step patterns

Step 4: Compare against dotforge