review-loop

review-loop performs a "work—review—revise" loop within the same session: after the agent completes a task, it launches a review sub-agent that evaluates only the output, scores it on a scale from 1 to 10, and provides specific feedback. The agent then revises based on this feedback. By default, the loop runs for at least two rounds and up to four rounds, requiring a score of 8 to pass, and concludes with a loop summary. SKILL.md requires that the review model must not be weaker than the working model, and specifies that this loop should not be used for simple edits, exploratory work, or tasks already covered by existing tests.

review-loop performs a "work—review—revise" loop within the same session: after the agent completes a task, it launches a review sub-agent that evaluates only the output, scores it on a scale from 1 to 10, and provides specific feedback. The agent then revises based on this feedback. By default, the loop runs for at least two rounds and up to four rounds, requiring a score of 8 to pass, and concludes with a loop summary. SKILL.md requires that the review model must not be weaker than the working model, and specifies that this loop should not be used for simple edits, exploratory work, or tasks already covered by existing tests.

Why it matters This turns "the first draft is not the final draft" into a fixed workflow, suitable for quality-critical tasks such as security, data pipelines, and financial logic.

skills.sh installs 112467 (+1202 / 24h)

Security Audit: Not audited

  • Not yet scanned

Rationale

  • skills.sh installs 1202 in 24 hours (indicating attention only)
  • Applicable and non-applicable scenarios are clearly documented; manual simulation methods are provided for platforms that do not support sub-agents

Usage

  • Say "implement X using review-loop" or "run review-loop on the file I just wrote"

Risks

  • Not yet audited; installation commands are not displayed; multiple review rounds can multiply token and time consumption

Open · Repository

Best Skills 每日数据(LinklyAI / @BlueeonY,CC BY 4.0)

Explore the latest AI Radar updates