Repo Analysis Skill

Analyze local repositories to find the best match for a work item based on context.

Process

1. List sibling repos: Find all directories with .git in REPOS_BASE_DIR
2. For each repo:
   • Pull latest from origin master/main
   • Read README.md and CLAUDE.md if they exist
   • Extract keywords and project descriptions
3. Score against context: Compare repo descriptions against:
   • Work item title
   • Work item description/body
   • PR URL hints (if any)
   • Project name from ADO
4. Output ranked list: Show repos sorted by confidence score

Usage

/repo-analysis <work-item-context>

Example:

/repo-analysis "Fix video player buffering issue - relates to player SDK"

Output Format

Repo Analysis Results:
1. video-player-sdk (95% match)
   - Keywords: video, player, SDK, streaming
   - Reason: Title mentions "player SDK"

2. media-services (60% match)
   - Keywords: media, transcoding, delivery
   - Reason: Related to video infrastructure

3. web-frontend (30% match)
   - Keywords: React, UI, components
   - Reason: May embed player

Implementation Steps

When invoked:

1. Read REPOS_BASE_DIR from environment or use default ../
2. List all directories, check for .git subdirectory
3. For each git repo:
   • git fetch origin (update refs)
   • Read README.md first 500 lines
   • Read CLAUDE.md if exists
   • Extract: project name, description, key technologies
4. Build keyword index per repo
5. Parse work item context for keywords
6. Score each repo (keyword overlap + description similarity)
7. Return top 5 matches with confidence scores