review-code
Multi-dimensional code review with structured reports. Analyzes correctness, readability, performance, security, testing, and architecture. Triggers on "review code", "code review", "审查代码", "代码审查".
58
67%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./.claude/skills/review-code/SKILL.mdQuality
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, well-crafted skill description that concisely communicates what the skill does (multi-dimensional code review across six specific dimensions with structured reports) and when to use it (explicit trigger terms in both English and Chinese). It uses third-person voice, avoids vague language, and provides enough specificity to distinguish it from general coding or debugging skills.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete dimensions of analysis: correctness, readability, performance, security, testing, and architecture. Also mentions structured reports as an output format. | 3 / 3 |
Completeness | Clearly answers 'what' (multi-dimensional code review with structured reports analyzing six dimensions) and 'when' (explicit triggers on specific phrases). The 'Triggers on' clause serves as an explicit 'Use when' equivalent. | 3 / 3 |
Trigger Term Quality | Includes natural English trigger terms ('review code', 'code review') and Chinese equivalents ('审查代码', '代码审查'). These are terms users would naturally say when requesting a code review. | 3 / 3 |
Distinctiveness Conflict Risk | The description carves out a clear niche: structured multi-dimensional code review. The specific trigger terms ('review code', 'code review') are distinct and unlikely to conflict with general coding assistance or debugging skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill provides a well-structured overview of a multi-dimensional code review process but suffers from significant verbosity and redundancy — the same phase structure is presented three times (architecture diagram, execution flow, reference table). The actual review logic is entirely deferred to external files that aren't provided, making the skill more of an index than actionable guidance. It would benefit greatly from condensing the overview and including at least one concrete example of performing a review and recording a finding.
Suggestions
Consolidate the architecture diagram, execution flow, and reference table into a single concise representation — currently the same information appears three times in different formats.
Add a concrete example showing what a single finding looks like (e.g., a code snippet with an identified issue, its severity classification, and the recommended fix) so Claude knows exactly what output to produce.
Remove explanations of concepts Claude already knows (what code review dimensions mean, what severity levels are) and instead focus on project-specific standards and thresholds.
Add validation checkpoints — e.g., after quick-scan, verify at least one risk area was identified; after deep-review, verify findings are classified correctly before generating the report.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extremely verbose with redundant information presented multiple times. The architecture diagram, execution flow, and reference documents table all repeat the same phase structure. The mandatory prerequisites section is overly dramatic with emoji warnings. Review dimensions and severity levels are described both inline and referenced to external specs, creating duplication. Much of this content (what code review dimensions are, what severity levels mean) is knowledge Claude already possesses. | 1 / 3 |
Actionability | The directory setup provides executable JavaScript/Bash code, and the execution flow gives a clear sequence. However, the actual review logic is entirely deferred to external files (phases/actions/*.md, specs/*.md) that are not provided. The skill tells Claude to read specs and follow templates but doesn't give concrete examples of how to actually perform a review or what a finding should look like. | 2 / 3 |
Workflow Clarity | The multi-step process is clearly sequenced (Phase 0 → collect-context → quick-scan → deep-review → generate-report → complete) with state outputs noted at each step. However, there are no validation checkpoints or feedback loops — no guidance on what to do if the quick scan finds nothing, if a dimension has no issues, or how to verify report quality before completing. | 2 / 3 |
Progressive Disclosure | The skill references many external files (specs, templates, phases, actions) with clear tables and links, which is good structure. However, since no bundle files are provided, all the actual actionable content lives in unreachable references. The SKILL.md itself is a verbose overview that repeats structural information rather than providing a lean summary with well-signaled pointers. The reference documents table at the end redundantly lists files already mentioned in the prerequisites section. | 2 / 3 |
Total | 7 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
allowed_tools_field | 'allowed-tools' contains unusual tool name(s) | Warning |
Total | 10 / 11 Passed | |
- Repository
- catlog22/Claude-Code-Workflow
- Commit
5ff5e86
- 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.