SKILL

Team Review

From Team Bundle by @jonathanhawkins

>-

This skill ships inside the team-bundle package. Install the package to get this skill plus everything else in the bundle.

sv install jonathanhawkins/team-bundle

Team Review — Parallel Code Review

Spawn an agent team where each reviewer examines code through a different lens, then the lead synthesizes findings.

Usage

/team-review [PR number or description of code to review]

How to Execute

Step 1: Determine Review Scope

  • If a PR number is given, fetch PR details with gh pr view <number> --json files,body,title
  • If a description or path is given, identify the files/changes to review
  • If nothing is specified, ask the user what to review

Step 2: Spawn Review Team

Create a team with 3 reviewers (add a 4th architecture reviewer for PRs touching 10+ files):

Security Reviewer:

You are the security reviewer. Review all changes for security vulnerabilities:
- Input validation and sanitization
- Authentication/authorization issues
- Injection vulnerabilities (SQL, XSS, command injection)
- Secrets or credentials in code
- OWASP Top 10 concerns
- Unsafe memory handling (for Rust: unsafe blocks, unchecked indexing)
Rate each finding as Critical/High/Medium/Low severity.
Send your findings to the lead when done.

Changes to review: [files/diff]

Performance Reviewer:

You are the performance reviewer. Review all changes for performance impact:
- Algorithm complexity (time and space)
- Database query efficiency (N+1, missing indexes)
- Memory leaks or excessive allocation
- Unnecessary computation or blocking operations
- Caching opportunities
- Hot path regressions
Rate each finding by impact: Critical/High/Medium/Low.
Send your findings to the lead when done.

Changes to review: [files/diff]

Correctness & Test Reviewer:

You are the correctness and test reviewer. Review all changes for:
- Logic errors and edge cases
- Test coverage of new/changed code paths
- Missing error handling
- Race conditions or concurrency issues
- API contract violations
- Whether tests actually test the right things (not just coverage theater)
Send your findings to the lead when done.

Changes to review: [files/diff]

Architecture Reviewer (optional, for large PRs):

You are the architecture reviewer. Evaluate:
- Does this change fit the existing architecture?
- Are abstractions appropriate (not over/under-engineered)?
- Are there better patterns for this use case?
- Will this be maintainable long-term?
- Does it introduce coupling that will be hard to unwind?
Send your findings to the lead when done.

Changes to review: [files/diff]

Step 3: Lead Synthesizes

After all reviewers report, the lead should:

  1. Deduplicate overlapping findings
  2. Prioritize by severity (Critical → Low)
  3. Present a unified review with actionable items
  4. Optionally post as a PR comment if the user requests it (gh pr comment)

Model Recommendations

  • Lead: Opus (synthesizes and prioritizes across reviewers)
  • Reviewers: Sonnet (focused analysis, saves tokens)

Example Invocations

/team-review 142
/team-review the authentication refactor in src/auth/
/team-review last 3 commits on this branch
/team-review engine-rs/crates/gdscene/ changes since last release