tools-repo-review（工具回购审查）

Repo Review Skill (Engineering Manager Edition)

This skill enables AI-powered repository analysis with three modes: contribution statistics, code quality assessment, and team/manager insights. Supports org-wide reviews, single repos, and saved repo groups.

Prerequisites

• Git CLI installed
• GitHub CLI (gh) for org-wide reviews
• Optional: Python 3.10+ with matplotlib for PNG chart generation

Quick Start

Review https://github.com/org/repo
Review the UP organization
Review my side-projects group

Progressive Onboarding

On first use, gather context through conversation:

Initial Questions

1. Role: Engineering Manager | Tech Lead | IC
2. Focus: Full review | Quick stats | Code quality only
3. Time period: 30 days | 90 days | YTD | Custom

Save preferences to config.yaml for future sessions.

Follow-up Questions (as needed)

• "I found 5 contributors. Focus on anyone specific?"
• "Compare against previous review?"
• "Flag contributors with < X commits?"

Analysis Modes

Mode 1: Contribution Statistics

Gather quantitative metrics using git commands:

# Clone repo to temp directory
git clone --quiet <repo_url> /tmp/repo_review_temp

# Contribution stats by author
git log --numstat --format="COMMIT|%H|%an|%ad" --date=short --since="2025-01-01" --all

# Weekly breakdown
git log --format="%ad" --date=short --since="2025-01-01" | sort | uniq -c

# Files touched by author
git log --author="Name" --name-only --pretty=format: | sort | uniq -c | sort -rn

# File type breakdown
git log --numstat --format="" | awk '{print $3}' | grep -oE '\.[^.]+$' | sort | uniq -c | sort -rn

Output metrics:

• Commits per author
• Lines added/deleted/net per author
• File type breakdown (frontend/backend/config/infra/docs)
• Weekly/monthly activity patterns

Mode 2: Code Quality Review

Examine the codebase for:

Architecture & Patterns

• Project structure organization
• Component/module boundaries
• DRY compliance
• Separation of concerns

Type Safety

• TypeScript strictness
• any usage
• Proper type exports

Error Handling

• Try/catch patterns
• Error boundaries
• Validation at boundaries

Database/API Patterns

• Query structure
• N+1 detection
• Input validation

Security

• Auth implementation
• Secret handling
• Input sanitization

Testing Indicators

• Test file presence
• Coverage indicators
• Test patterns

Mode 3: Team/Manager Assessment

Combine quantitative data with qualitative code review:

Per-Contributor Analysis:

• Volume and consistency
• Scope of work (full-stack vs narrow)
• Quality signals
• Growth indicators

Team Health:

• Bus factor
• Knowledge silos
• Cross-functional work
• Velocity trends

Review Targets

1. Single Repository

Review https://github.com/org/repo
Review https://github.com/org/repo --since 2025-06-01

2. Organization-Wide

Review the UP organization
Review all repos in UP org

Uses GitHub CLI:

gh repo list UP --limit 100 --json name,sshUrl,url

3. Saved Groups

Review my side-projects group
Review client-work group

4. Managing Groups

Add https://github.com/user/repo to side-projects
Create a new group called "experiments"
Remove old-project from experiments
Show all my saved groups

Visualization Options

Mermaid Diagrams (for markdown reports)

Contribution Timeline:

gantt
    title Contribution Timeline
    dateFormat YYYY-MM-DD
    section Contributor1
    Feature A     :2025-12-30, 2d
    Feature B     :2026-01-02, 1d
    section Contributor2
    Bug Fix       :2026-01-01, 1d

Work Breakdown:

pie title Work Breakdown
    "Frontend" : 45
    "Backend" : 30
    "Config" : 15
    "Docs" : 10

ASCII Charts (for terminal/quick view)

Author          | Net LoC | Contribution
----------------|---------|----------------------------------
Alice           |   2,872 | ████████████████████████████ 100%
Bob             |   1,020 | ████████████░░░░░░░░░░░░░░░░  35%
Charlie         |      18 | █░░░░░░░░░░░░░░░░░░░░░░░░░░░   1%

Weekly Trend:

Week 1: ████████████ 1,200 LoC
Week 2: ███████████████ 1,450 LoC ↑
Week 3: █████████ 890 LoC ↓ ⚠️

Script Charts (for presentations)

Run scripts/repo_review_om.py for matplotlib PNG generation:

python scripts/repo_review_om.py --repo https://github.com/org/repo --chart

Manager-Focused Outputs

Performance Review Language

Strong Performer:

"[Name] demonstrated strong capabilities, delivering [X] features across [Y] commits totaling [Z] net lines. Notable contributions include [specific feature]. Their work showed [quality indicator - e.g., proper error handling, TypeScript usage, test coverage]."

Solid Contributor:

"[Name] contributed consistently with [X] commits and [Y] net lines. Primary focus was [area]. Recommend expanding into [adjacent area] for growth."

Needs Attention:

"[Name] had limited participation with [X] commits totaling [Y] lines. Contributions were confined to [narrow scope]. Recommend 1:1 to understand blockers and assign stretch goals."

1:1 Conversation Starters

Growth/Promotion Readiness

Junior → Mid-Level Signals:

• Works across frontend AND backend
• Writes database migrations/schema changes
• Fixes bugs introduced by others (not just own code)
• Adds tests without being prompted
• Clear commit messages with context
• Participates in code review

Mid → Senior Signals:

• Designs features end-to-end
• Mentors juniors (visible in PR reviews)
• Identifies and addresses technical debt
• Documents architectural decisions
• Considers security implications
• Optimizes for maintainability

Red Flag Detection

Team Health Metrics

Velocity Trend:

Period      | Team LoC | Trend    | Notes
------------|----------|----------|------------------
Week 1      |    1,200 | baseline |
Week 2      |    1,450 | ↑ 21%    | Sprint push
Week 3      |      890 | ↓ 39%    | ⚠️ Investigate
Week 4      |    1,100 | ↑ 24%    | Recovery

Knowledge Distribution:

• Bus Factor: How many people can modify each critical area?
• Silo Score: % of code touched by only one person
• Cross-functional Index: % of contributors working across stack

Contribution Balance:

Ideal: Relatively balanced contributions
⚠️ Warning: Top contributor has >70% of commits
🚨 Critical: One person has >90% of commits

Configuration

The skill uses config.yaml for persistent settings. Created on first use.

Config Structure

# Default analysis settings
defaults:
  since: "2025-01-01"
  analysis_modes:
    - contributions
    - quality
    - team
  output_format: markdown  # markdown | ascii | json
  role: engineering_manager  # engineering_manager | tech_lead | ic

# Saved organizations
orgs:
  up:
    github_org: "UP"
    description: "UP Company repositories"
    active: true
    contributors:
      - name: "Alice Smith"
        aliases: ["alice", "asmith"]
        role: "Senior Engineer"
        active: true
      - name: "Bob Jones"
        aliases: ["bob", "bjones"]
        role: "Junior Engineer"
        active: true
      - name: "Former Employee"
        aliases: ["former"]
        role: "Engineer"
        active: false  # No longer with team
    exclude_authors:
      - "dependabot[bot]"
      - "github-actions[bot]"
      - "renovate[bot]"

# Saved repo groups (for non-org collections)
groups:
  side-projects:
    description: "Personal weekend experiments"
    active: true
    repos:
      - "https://github.com/user/project-a"
      - "https://github.com/user/project-b"

  client-work:
    description: "Freelance client projects"
    active: true
    repos:
      - "https://github.com/client/app"

  archived:
    description: "Old projects for reference"
    active: false
    repos:
      - "https://github.com/user/old-thing"

# Session tracking
last_session:
  target: "up"
  target_type: "org"  # org | group | repo
  date: "2026-01-10"

Managing Config

Add contributor:

Add contributor "New Person" with alias "newperson" to UP org

Deactivate contributor:

Mark "Former Employee" as inactive in UP org

Update org:

Set UP org as inactive

Example Workflows

Weekly Team Standup Prep

User: "Quick stats on UP org for the past week"

AI: [Fetches last 7 days, shows commit counts per active contributor, flags anyone with 0 commits]

Monthly Review

User: "Full review of UP org for December"

AI: [Comprehensive analysis with charts, per-contributor breakdown, team health metrics, comparison to November if available]

1:1 Preparation

User: "Prep 1:1 notes for Bob based on UP contributions"

AI: [Pulls Bob's stats across all org repos, correlates with code quality observations, generates talking points and growth recommendations]

New Hire Onboarding Check

User: "How is the new junior doing in UP org? Started 2 weeks ago."

AI: [Filters to last 2 weeks, compares to typical onboarding velocity, identifies areas of focus, suggests pairing opportunities]

Quarterly Summary

User: "Generate Q4 summary for UP team"

AI: [Aggregates Oct-Dec data, identifies top contributors, flags concerns, generates executive summary suitable for leadership]

Project Kickoff

User: "Add https://github.com/client/new-project to client-work and give me initial review"

AI: [Adds to group, clones repo, establishes baseline metrics, identifies current contributors and code state]

Troubleshooting

GitHub CLI Not Authenticated

Error: gh: not logged in

Fix: Run `gh auth login` and follow prompts

Private Repo Access

Error: Repository not found (404)

Fix: Ensure gh CLI has access, or use SSH URL if you have keys configured

Large Org Timeout

For orgs with 50+ repos, consider:
1. Running with --limit flag
2. Excluding archived repos
3. Focusing on specific repos via group

Git Config Issues (Multiple Authors)

Contributor showing as multiple people:

Fix: Have them run:
  git config --global user.name "Consistent Name"
  git config --global user.email "consistent@email.com"

Workaround: Add all variations to aliases in config.yaml

Script Reference

For advanced chart generation, use the included Python script:

cd skills/repo-review/scripts

# Single repo with chart
python repo_review_om.py --repo https://github.com/org/repo --chart

# Org-wide
python repo_review_om.py --all --dir /tmp/org-repos

# Date range
python repo_review_om.py --repo URL --since 2025-06-01 --until 2025-12-31

# Exclude docs
python repo_review_om.py --repo URL --no-docs

Output: CSV data file + PNG chart (if --chart flag used)