auto-review-loop
Autonomous multi-round research review loop. Repeatedly reviews using Claude Code via claude-review MCP, implements fixes, and re-reviews until positive assessment or max rounds reached. Use when user says "auto review loop", "review until it passes", or wants autonomous iterative improvement.
72
88%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a strong skill description that clearly communicates a specific autonomous review loop workflow, includes explicit trigger terms, and answers both what the skill does and when to use it. The description is concise, uses third person voice, and carves out a distinct niche that would be easily distinguishable from simpler review or code analysis skills.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: 'reviews using Claude Code via claude-review MCP', 'implements fixes', 're-reviews until positive assessment or max rounds reached'. The description clearly articulates the multi-step autonomous loop process. | 3 / 3 |
Completeness | Clearly answers both 'what' (autonomous multi-round research review loop that reviews, implements fixes, and re-reviews) and 'when' (explicit 'Use when' clause with specific trigger phrases like 'auto review loop' and 'review until it passes'). | 3 / 3 |
Trigger Term Quality | Includes natural trigger terms users would say: 'auto review loop', 'review until it passes', 'autonomous iterative improvement'. These are realistic phrases a user would use when requesting this specific workflow. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive with a clear niche: autonomous iterative review loop using claude-review MCP. The combination of autonomous looping, MCP tool usage, and iterative fix-and-review cycle makes it unlikely to conflict with simple code review or one-shot review skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured, highly actionable skill for an autonomous review loop with clear multi-phase workflow, concrete MCP tool calls, and robust state persistence for crash recovery. Its main weakness is length — some content is repeated (MCP polling instructions appear twice) and the monolithic structure could benefit from splitting detailed templates into referenced files. The conditional features (human checkpoint, Feishu notifications) add complexity but are well-gated.
Suggestions
Deduplicate the MCP polling instructions — define the poll-until-done pattern once and reference it in both Round 1 and Round 2+ sections.
Consider extracting the prompt templates and JSON schemas into separate reference files to reduce the main skill's length and improve progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is fairly long (~200+ lines) with some redundancy — the MCP polling instructions are repeated nearly verbatim for round 1 and round 2+, and some sections (like the Feishu notification and human checkpoint) add conditional complexity that could be more tightly structured. However, it mostly avoids explaining concepts Claude already knows and each section serves a purpose. | 2 / 3 |
Actionability | The skill provides concrete MCP tool calls with specific parameter formats, exact JSON schemas for state persistence, precise prompt templates for the reviewer, detailed parsing instructions for assessments, and specific file paths. The workflow is fully executable with copy-paste ready templates and clear tool invocation patterns. | 3 / 3 |
Workflow Clarity | The multi-phase workflow (A→B→C→D→E) is clearly sequenced with explicit stop conditions, validation checkpoints (score threshold parsing), error recovery (state persistence for compaction recovery), and feedback loops (review→fix→re-review). The initialization section handles multiple edge cases (fresh start, resume, stale state) with clear decision logic. | 3 / 3 |
Progressive Disclosure | The skill references shared protocols (output-versioning.md, output-manifest.md, output-language.md) which is good progressive disclosure, but the main SKILL.md itself is quite long and monolithic. The prompt templates, state JSON schema, and review document format could potentially be split into separate reference files. No bundle files are provided to verify the referenced paths exist. | 2 / 3 |
Total | 10 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Commit
a425a71
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.