study-repo

---

name: study-repo description: Deep analysis of an external repository's architecture, patterns, and design decisions. Generates comprehensive influence document and updates improvements.md with prioritized recommendations. agent: general-purpose allowed-tools: Read, Grep, Glob, Bash disable-model-invocation: false argument-hint: [repo-path] [--focus=aspect]

Repository Study Skill

Deep analysis of external repositories to identify patterns, architecture decisions, and opportunities for adoption in ClaudeMemory.

Prerequisites

Before running this skill, clone the repository you want to study:

# For public repos
git clone https://github.com/user/project /tmp/study-repos/project-name

# For private repos (requires gh CLI)
gh repo clone user/project /tmp/study-repos/project-name

# Use --depth 1 for speed (recommended)
git clone --depth 1 https://github.com/user/project /tmp/study-repos/project-name

Then invoke: /study-repo /tmp/study-repos/project-name

Optional Focus Mode: Narrow analysis to specific aspect

/study-repo /tmp/study-repos/project-name --focus="MCP implementation"
/study-repo /tmp/study-repos/project-name --focus="testing strategy"

See .claude/skills/study-repo/focus-examples.md for more examples.

CRITICAL: Memory Discipline (no external-tech misattribution)

When studying an external repo you will read its README, gemspec, and source — and you will see things like "uses Postgres", "runs on AWS", "built with Rails". These are facts about the external project, not about this project.

Do NOT call memory.store_extraction with the external project's tech stack as uses_database / uses_framework / uses_language / deployment_platform / auth_method predicates. That misattribution caused 27 facts to be stored about ClaudeMemory in the 2026-04-23/24 window that all had to be hand-rejected (see improvements.md #61, quality_review.md 2026-04-30 note). The corpus damage was real even though the cleanup worked — every misattributed fact takes a round trip through the database, conflict-detection, and the user's claude-memory reject queue.

The rule. While /study-repo is running, the only memory.store_extraction calls allowed are:

• predicate=reference for descriptions of the external project ("X is a plugin/library/CLI that…"). The dashboard's Knowledge → References panel is the right home for these.
• Facts genuinely about this project ClaudeMemory that you derive from contrast with the studied repo (e.g., a decision: "Adopt RRF fusion from QMD because…"). These belong as decision / convention / architecture with subject=repo or subject=claude_memory AND a reason clause embedded.

The hard ban. Any single-value cardinality predicate (uses_database, deployment_platform, auth_method) populated with the studied project's tech is forbidden. If in doubt, write the observation into the influence document (docs/influence/<project>.md) — that file IS the right artifact for "what does this external project use" — and skip memory.store_extraction entirely.

If the user asks "did the studied project use X?" later, the answer lives in docs/influence/, not in memory facts.

Analysis Phases

Follow these phases systematically to ensure comprehensive coverage:

Phase 1: Repository Context (10 minutes)

Goal: Understand project purpose, maturity, and ecosystem

Tasks:

1. Read README.md, CHANGELOG.md, LICENSE
2. Examine gemspec/package.json for dependencies and metadata
3. Check GitHub metadata (stars, issues, PRs, contributors)
4. Review CI configuration (.github/workflows/, .gitlab-ci.yml)
5. Check documentation structure (docs/, wiki)

Output: Executive Summary section with:

• Project purpose (1-2 sentences)
• Key innovation (what makes it unique)
• Technology stack table
• Production readiness assessment

Phase 2: Architecture Mapping (15 minutes)

Goal: Map high-level structure and module organization

Tasks:

1. Run tree -L 3 -d to understand directory structure
2. Identify entry points (bin/, exe/, main files)
3. Map core modules and their responsibilities
4. Identify data model (database schema, structs, classes)
5. Document dependencies between modules

Output: Architecture Overview section with:

• Data model description
• Design patterns identified
• Module organization diagram (text-based)
• Comparison table vs ClaudeMemory

Phase 3: Pattern Recognition (20 minutes)

Goal: Identify key design patterns and conventions

Tasks:

1. Search for common patterns:
   • Dependency injection
   • Command pattern
   • Repository pattern
   • Factory pattern
   • Observer pattern
2. Review naming conventions
3. Check error handling approaches
4. Examine configuration management
5. Identify testing patterns

Output: Key Components Deep-Dive with:

• Component purpose
• Code examples with file:line references
• Design decision rationale

Phase 4: Code Quality Assessment (15 minutes)

Goal: Evaluate testing, documentation, and performance

Tasks:

1. Analyze test coverage and structure
2. Review documentation quality
3. Check for performance optimizations
4. Examine logging and observability
5. Look for code quality tools (linters, formatters)

Output: Comparative Analysis section with:

• What they do well (with evidence)
• What we do well (our advantages)
• Trade-offs table

Phase 5: Comparative Analysis (15 minutes)

Goal: Compare their approach with ClaudeMemory's

Tasks:

1. For each major component, compare:
   • Their implementation
   • Our implementation
   • Pros/cons of each approach
2. Identify areas where they excel
3. Identify areas where we excel
4. Document fundamental architectural differences

Output: Enhanced Comparative Analysis with:

• Side-by-side comparison
• Trade-off analysis
• Context for adoption decisions

Phase 6: Adoption Opportunities (20 minutes)

Goal: Extract actionable recommendations with priorities

Tasks:

1. List all potential adoptions
2. For each, assess:
   • Value (quantified benefit)
   • Effort (implementation complexity)
   • Risk (compatibility, maintenance)
   • Fit (alignment with our architecture)
3. Prioritize as High ⭐ / Medium / Low
4. Identify features to avoid
5. Propose implementation phases

Output: Adoption Opportunities section with:

• High Priority ⭐ items (adopt soon)
• Medium Priority items (consider)
• Low Priority items (defer)
• Features to Avoid (with reasoning)
• Implementation Recommendations (phased plan)

Output Format

Generate a comprehensive influence document following this structure (see .claude/skills/study-repo/analysis-template.md for full template):

# [Project Name] Analysis

*Analysis Date: YYYY-MM-DD*
*Repository: GitHub URL*
*Version/Commit: tag or SHA*

---

## Executive Summary
[Project purpose, key innovation, tech stack, maturity]

## Architecture Overview
[Data model, design patterns, module organization, comparison table]

## Key Components Deep-Dive
[Component 1, Component 2, etc. with code examples and file:line references]

## Comparative Analysis
[What they do well, what we do well, trade-offs]

## Adoption Opportunities

### High Priority ⭐
#### 1. [Feature]
- **Value**: [Benefit]
- **Evidence**: [file:line proof]
- **Implementation**: [Approach]
- **Effort**: [Estimate]
- **Trade-off**: [Costs]
- **Recommendation**: ADOPT / CONSIDER / DEFER

[Medium and Low sections follow]

### Features to Avoid
[With reasoning]

## Implementation Recommendations
[Phase 1, Phase 2, etc.]

## Architecture Decisions
[What to preserve, adopt, reject]

## Key Takeaways
[Main learnings and next steps]

Critical Requirements:

• All claims must include file:line references
• Code examples for key patterns
• Priority markers (⭐) on High Priority items
• Comparison table vs ClaudeMemory
• Quantified benefits where possible
• Clear ADOPT/CONSIDER/DEFER recommendations

Integration with improvements.md

After creating the influence document, update docs/improvements.md:

1. Extract High Priority items (⭐ marked) from influence doc
2. Format as new section:

   ## [Project Name] Study (YYYY-MM-DD)
   
   Source: docs/influence/[project-name].md
   
   ### High Priority Recommendations
   
   - [ ] **[Feature 1]**: [Brief description]
     - Value: [Quantified benefit]
     - Evidence: [file:line from their repo]
     - Effort: [Estimate]
   
   - [ ] **[Feature 2]**: [...]
   

3. Append to improvements.md preserving existing structure
4. Update "Last updated" timestamp at top of file

Success Criteria

The analysis is complete when all of these are verified:

• ✅ docs/influence/<project_name>.md created with all sections
• ✅ Executive Summary includes project metadata and tech stack
• ✅ Architecture Overview has comparison table vs ClaudeMemory
• ✅ Key Components include code examples with file:line references
• ✅ All major claims have evidence citations
• ✅ Adoption Opportunities use ⭐ for High Priority items
• ✅ Each opportunity has Value/Evidence/Implementation/Effort/Trade-off
• ✅ Clear ADOPT/CONSIDER/DEFER recommendations given
• ✅ Implementation Recommendations organized in phases
• ✅ Trade-offs section discusses costs and risks
• ✅ docs/improvements.md updated with new dated section
• ✅ High priority items extracted and formatted as tasks
• ✅ Existing improvements.md structure preserved
• ✅ Console summary shows key findings

Focus Mode

When --focus flag is provided, narrow the analysis:

1. Parse focus topic from arguments
2. Filter files to only relevant ones:
   • For "testing": test/, spec/, .github/workflows/
   • For "MCP": mcp/, server/, tools/
   • For "database": schema, migrations, queries
   • For "CLI": bin/, exe/, commands/
3. Narrow exploration to focused aspect
4. Target recommendations to focused area
5. Faster execution with reduced scope

See .claude/skills/study-repo/focus-examples.md for detailed examples.

Error Handling

Handle common error cases gracefully:

• Invalid path: Display clear error with example usage
• Empty directory: Warn about missing files
• No README: Note absence but continue analysis
• Large repo: Suggest using --focus flag
• Permission issues: Check file permissions and suggest fixes

Execution Steps

1. Parse arguments: Extract repo path and optional focus
2. Validate path: Check directory exists and is readable
3. Detect project name: Extract from path or .git/config
4. Run analysis phases: Follow Phase 1-6 systematically
5. Generate influence doc: Use template with collected data
6. Update improvements.md: Extract and append High Priority items
7. Display summary: Show key findings and output locations
8. Exit with success: Return 0 if complete

Integration with /improve

The recommendations added to docs/improvements.md can be implemented using /improve:

# Study external project
/study-repo /tmp/study-repos/some-project

# Review recommendations
cat docs/influence/some-project.md

# Implement selected improvements
/improve

This creates a complete workflow: /study-repo → recommendations → /improve → implementation