code-reviewer-chorus

Reviews aggregate code changes across all tasks of a Chorus Idea and posts a structured verdict comment.

1.2k|104|Updated Feb 4, 2026
One-click install
npx skills add https://github.com/Chorus-AIDLC/Chorus --skill code-reviewer-chorus
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-reviewer-chorus
Source: https://github.com/Chorus-AIDLC/Chorus/tree/main/public/skill/code-reviewer-chorus
Command: npx skills add https://github.com/Chorus-AIDLC/Chorus --skill code-reviewer-chorus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Per-task code reviews can each pass while the overall feature is broken. This Skill performs the final read-only adversarial review of an entire Chorus Idea's aggregate code change, catching cross-task integration failures, architecture drift, security holes, and regressions that isolated reviews structurally miss.

Core Features & Use Cases

  • Aggregate Feature Review: Fetches the Idea, approved proposals, documents, and tasks via Chorus MCP tools, infers the diff scope from task work reports, and reviews the whole feature across six dimensions including integration, security, and test coverage.
  • Read-Only Verification: Runs the project's build, test, and lint commands with a strict read-only Bash policy, treating any build or test failure as an automatic FAIL verdict.
  • Structured Verdict Contract: Classifies findings as BLOCKER or NOTE and posts a single comment ending in an exact literal verdict (VERDICT: PASS, PASS WITH NOTES, or FAIL) that automation greps for.
  • Use Case: After all tasks of an Idea pass individual review, run this Skill to verify the tasks actually integrate, then let the orchestrator ship on PASS or spawn fix tasks on FAIL.

Quick Start

Ask the agent to run the code-reviewer-chorus skill on a given Idea UUID to review the whole feature's aggregate code change and post its verdict comment.

Frequently Asked Questions about code-reviewer-chorus

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review an entire feature across multiple tasks in Chorus?

Invoke the code-reviewer-chorus skill with an Idea UUID. It fetches the Idea, approved proposals, documents, and tasks via Chorus MCP tools, infers the aggregate diff from task work reports, and posts one structured verdict comment on the Idea.

What does the code reviewer check that per-task review misses?

It checks cross-task integration and contract consistency, architecture drift, security holes visible only in combination, regressions in code no single task owned, feature-level test coverage gaps, and overall correctness of the aggregate change.

Can the code reviewer modify files or run git commands?

No. It is strictly read-only: no file creation or modification, no package installs, and no git write operations. Its only side effect is posting one comment via chorus_add_comment; Bash is limited to test, build, lint, and read-only inspection commands.

What verdicts does the code review produce?

It ends its comment with exactly one literal: VERDICT: PASS, VERDICT: PASS WITH NOTES, or VERDICT: FAIL. Any BLOCKER finding forces FAIL, only NOTEs yield PASS WITH NOTES, and automation greps for these exact strings.

How does round 2 of a code review differ from round 1?

Round 2+ focuses only on whether previous BLOCKERs were fixed, re-reading just the tied files and re-running only the specific commands. It does not introduce new NOTEs, re-scan unrelated code, or rerun the full suite.

What happens after a FAIL verdict in Chorus?

The reviewer stays read-only. The orchestrator invokes Quick Dev to create fix tasks on the original approved proposal, and the reviewer is re-run only after all fix tasks are done; a failed or cancelled fix stops the loop and escalates.