task-reviewer-chorus

Verifies Chorus task implementations against acceptance criteria 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 task-reviewer-chorus
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: task-reviewer-chorus
Source: https://github.com/Chorus-AIDLC/Chorus/tree/main/public/skill/task-reviewer-chorus
Command: npx skills add https://github.com/Chorus-AIDLC/Chorus --skill task-reviewer-chorus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

LLM-written code often passes its own circular tests while silently diverging from acceptance criteria and proposal documents, leaving human reviewers to manually re-verify every submitted task.

Core Features & Use Cases

  • Independent AC Verification: Fetches the task, acceptance criteria, and proposal documents via Chorus MCP tools, then checks each criterion against actual code and test evidence.
  • Read-Only Adversarial Review: Runs only test/build/lint and read-only inspection commands, classifying findings as BLOCKER or NOTE without ever modifying the project.
  • Structured Verdict Output: Posts a single concise comment ending in an exact machine-greppable verdict: VERDICT: PASS, PASS WITH NOTES, or FAIL.
  • Use Case: A developer agent marks a Chorus task complete; this skill independently reruns the test suite, cross-references the PRD, probes edge cases, and posts a verdict the admin uses to verify or reopen the task.

Quick Start

Ask the agent to review the submitted Chorus task with a given taskUuid and post its verdict comment.

Frequently Asked Questions about task-reviewer-chorus

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

FAQPage Schema
How do I review a Chorus task implementation automatically?

Provide the taskUuid to the task-reviewer-chorus skill. It fetches the task, acceptance criteria, and proposal documents via Chorus MCP tools, runs the project's test and build commands, verifies each criterion, and posts one structured VERDICT comment on the task.

What verdicts does the Chorus task reviewer post?

It posts exactly one of three literal strings: VERDICT: PASS when no findings exist, VERDICT: PASS WITH NOTES when only non-blocking notes exist, and VERDICT: FAIL when any BLOCKER finding exists. Automation greps for these exact strings.

Can the task reviewer modify code or fix issues it finds?

No. The skill is strictly read-only: it cannot create, modify, or delete files, install packages, or run git write operations. Its only side effect is posting a single comment via chorus_add_comment.

What is the difference between a BLOCKER and a NOTE finding?

A BLOCKER blocks correctness: unmet acceptance criteria, build or test failures, or divergence from proposal documents. A NOTE is non-blocking: pseudocode signature mismatches, wording differences, style suggestions, or hallucination-risk details like SDK versions and CLI flags.

How does round 2 re-review work in the Chorus review workflow?

In round 2 and later, the reviewer checks only whether previous BLOCKERs were fixed, re-reading the specific files and re-running only the tied tests. It does not introduce new NOTEs or re-scan unrelated code, and passes the task if all prior blockers are resolved.