From plugin-management
Use when auditing or refactoring a reusable agent artifact such as a `SKILL.md` bundle or standalone agent definition for trigger quality, bundle structure, metadata parity, progressive disclosure, or reusable capability fit.
How this skill is triggered — by the user, by Claude, or both
Slash command
/plugin-management:skill-reviewerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review reusable agent artifacts as bundles of files, not as essay prompts.
agents/openai.yamlexamples/good-file-review.jsonexamples/good-synthesized-report.mdexamples/needs-improvement-file-review.jsonexamples/rethink-synthesized-report.mdreferences/anti-patterns.mdreferences/artifact-adapters.mdreferences/finding-to-advice.mdreferences/reviewer-json.mdreferences/runtime-targets.mdreferences/subagent-patterns.mdreferences/test-cases.mdreferences/topology-rules.mdReview reusable agent artifacts as bundles of files, not as essay prompts.
SKILL.mdreferences/, examples/, runtime metadata, or bundled scriptsDo not use this skill for generic prompt review, slash commands, or broad documentation audits with no reusable agent behavior.
Use references/skill-standards/README.md as the authoring baseline. Local reviewer references map those shared rule IDs to finding kinds, priorities, and output expectations.
Use references/runtime-targets.md as the source of truth for host-specific definition files.
This skill does not use weighted scoring.
Run this workflow in order for each artifact.
SKILL.md, the primary definition file is SKILL.md.agents/openai.yaml as required runtime metadata when the artifact depends on it, even if it is not linked from SKILL.md.Topology Reviewer.
references/topology-rules.md to apply DISCLOSURE-01, TOPOLOGY-01, TOPOLOGY-02, and METADATA-01.references/reviewer-json.md.File Reviewer passes on the primary file and any load-bearing support files.
references/artifact-adapters.md to map shared rule IDs to role-specific expectations.references/anti-patterns.md for finding kinds and priority calibration.references/finding-to-advice.md.good, needs_improvement, or rethink.Use narrow reviewers only when they reduce complexity. See references/subagent-patterns.md.
Input:
Checks:
Output:
role must be topologyInput:
Roles:
primaryreferenceexamplemetadatasubagent for standalone agent definition filesChecks:
Output:
Input:
Responsibilities:
The synthesizer is the only reviewer allowed to decide whether the artifact deserves to exist as a reusable capability.
Default to these user-facing sections:
StatusKey FindingsWhat To KeepRefactor PlanStatus values:
good: only minor P3 issues or no material issuesneeds_improvement: worthwhile artifact with fixable issuesrethink: structurally misleading, too fragmented, or not worth keeping in its current shapeDefault to completing the review without interruption.
Ask a user question only when a missing fact would materially change:
Otherwise make a reasonable assumption, state it briefly, and continue.
Load these files as needed:
../../references/skill-standards/README.mdreferences/anti-patterns.mdreferences/artifact-adapters.mdreferences/finding-to-advice.mdreferences/reviewer-json.mdreferences/runtime-targets.mdreferences/subagent-patterns.mdreferences/test-cases.mdreferences/topology-rules.mdnpx claudepluginhub moonsong-labs/knowledge-work-plugins --plugin plugin-managementReviews and improves AI agent skills (SKILL.md files) against best practices, scoring 10 quality dimensions and rewriting problem areas.
Reviews Claude Code SKILL.md files and agent definitions for frontmatter validity, token budget, description quality, tool scope, and agent contract compliance. Loads on skill or agent files.
Audits SKILL.md files against the AgentOps template with a three-pass process: structural hygiene via heal-skill, 8 content-discipline checks, and an advisory 0-30 productization rubric. Read-only; reports findings without modifying files.