code-review
SkillAI & modelsCode review the current branch against origin/master using 7 parallel review agents with confidence scoring.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the code-review skill
What this skill tells your AI
The instructions your AI receives, as published by the-agency-ai/the-agency in .claude/skills/code-review/SKILL.md and read by ahel’s review.
Code Review
Code review the current branch against origin/master using 7 parallel review agents with confidence scoring. Runs locally against git diff; no GitHub PR required.
Arguments
- $ARGUMENTS: Optional branch name, PR number, or empty for current branch.
Steps
Step 1: Determine branch and diff
- If PR number provided:
gh pr view {number} --json headRefNameto get branch - If branch name provided: use it directly
- If empty: use current branch via
git rev-parse --abbrev-ref HEAD - Compute diff:
git diff origin/master...{branch}
Step 2: Gather CLAUDE.md files
Read all CLAUDE.md files in the repo to understand project conventions.
Step 3: Summarize changes
Run git diff --stat origin/master...{branch} and categorize changed files by area.
Step 4: Launch 7 parallel review agents
Launch all 7 as Sonnet agents in a single parallel call:
- CLAUDE.md compliance — Check changes against project conventions
- Bug scan — Logic errors, null handling, type mismatches, runtime crashes
- Git history —
git logcontext, are changes consistent with commit messages? - Prior review context — Check for existing review files, are prior findings addressed?
- Code comments — Misleading comments, missing docs for public APIs
- Test coverage — Coverage gaps, missing edge cases, stale assertions
- Test consistency — Do tests match implementation? Stale mocks?
Step 5: Score findings
Send all findings to a reviewer-scorer agent (model: haiku) for confidence scoring (0-100).
Step 6: Filter
Remove findings with score < 80. Deduplicate.
Step 7: Save review
Derive the project name from the branch (last segment after /, or the branch name itself).
Save to agency/workstreams/{workstream}/code-reviews/{workstream}-review-{YYYYMMDD-HHMM}.md where {workstream} is derived from the branch name (segment before /).
Step 8: Post to GitHub (if PR exists)
If a PR exists for the branch, offer to post comments via gh pr review.
Step 9: Present summary
Show the review results: total findings, breakdown by category, high-confidence issues.
Signals
- GitHub stars
- 55
- Forks
- 8
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
code-review-the-agency-ai- Source
- github.com/the-agency-ai/the-agency