molecule-core/plugins/molecule-dev/skills/review-loop/SKILL.md
Hongming Wang 24fec62d7f initial commit — Molecule AI platform
Forked clean from public hackathon repo (Starfire-AgentTeam, BSL 1.1)
with full rebrand to Molecule AI under github.com/Molecule-AI/molecule-monorepo.

Brand: Starfire → Molecule AI.
Slug: starfire / agent-molecule → molecule.
Env vars: STARFIRE_* → MOLECULE_*.
Go module: github.com/agent-molecule/platform → github.com/Molecule-AI/molecule-monorepo/platform.
Python packages: starfire_plugin → molecule_plugin, starfire_agent → molecule_agent.
DB: agentmolecule → molecule.

History truncated; see public repo for prior commits and contributor
attribution. Verified green: go test -race ./... (platform), pytest
(workspace-template 1129 + sdk 132), vitest (canvas 352), build (mcp).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-13 11:55:37 -07:00

3.1 KiB

name description
review-loop Orchestrate a multi-round implementation + review cycle. Use when coordinating a feature that requires implementation (FE/BE), design review (UIUX), security review, and QA verification. Ensures QA findings get routed back for fixes until clean.

Review Loop

Orchestrate implementation through multiple rounds until QA reports zero issues. This prevents the one-shot delegation problem where QA finds bugs but nobody fixes them.

When to Use

Use this when you're a coordinator (Dev Lead, PM) assigning a feature that involves multiple specialists.

The Loop

Round 1: Design + Implementation (parallel where possible)

  1. Identify all stakeholders — before delegating, ask: who needs to be involved?

    • UI work → UIUX Designer reviews interaction design FIRST
    • Credentials / auth / secrets → Security Auditor reviews
    • API changes → Backend Engineer + Frontend Engineer coordinate
    • Everything → QA Engineer is the final gate
  2. Delegate design review first (if UI work):

    delegate_task_async → UIUX Designer: "Review the interaction design for [feature]"
    
  3. Delegate implementation (after design, or parallel if non-UI):

    delegate_task_async → Frontend Engineer: "Implement [feature] following UIUX spec"
    delegate_task_async → Backend Engineer: "Add [endpoint]" (if needed)
    delegate_task_async → Security Auditor: "Review [feature] for [specific concerns]"
    
  4. Delegate QA (can start in parallel — QA reads existing code while FE works):

    delegate_task_async → QA Engineer: "Review [feature], run full test suite, write missing tests, grep for convention violations"
    
  5. Collect all results via check_task_status on each delegation.

Round 2: Fix QA Findings (if any issues found)

If QA reported issues:

  1. Send QA's findings back to the implementer:

    delegate_task → Frontend Engineer: "QA found these issues in your implementation:
    [paste QA's specific findings with file:line references]
    Fix all of them and report back."
    
  2. Re-run QA on the fixes:

    delegate_task → QA Engineer: "FE applied fixes for your findings. Re-verify:
    [paste the specific issues that were fixed]
    Run the test suite again. Report if any issues remain."
    
  3. If QA still finds issues → repeat Round 2.

Round 3: Final Sign-off

When QA reports zero issues:

  • Compile the full report: what was implemented, what was fixed, test results
  • Report to PM / CEO with substance, not just "done"

Key Rules

  • Never skip QA. Even if FE says "I tested it." QA verifies independently.
  • Never skip Security for credential-related features. A secrets panel without security review is a liability.
  • QA findings are not optional. If QA found it, it gets fixed. Period.
  • Use parallel delegation. delegate_task_async to all specialists at once, then collect with check_task_status. Don't serialize what can be concurrent.
  • Ask side questions. If FE needs to know the API shape, FE should delegate_task directly to BE — don't relay through the lead.