An exceptionally lean and actionable process document: exact commands, a deterministic impact-leveling algorithm, explicit evidence rules, and a complete output template, with validation loops built into the workflow. Its only structural weakness is that all detail lives inline in SKILL.md and the external surfaces.md dependency is never located.

Dimension	Reasoning	Score
Conciseness	The body is dense throughout — compressed tables, one-line judgment rules, a concrete historical case (#3492), and zero padding — and its only explanatory paragraph is repo-specific regression history Claude cannot infer. Every section carries operational content; nothing explains concepts Claude already knows.	5 / 5
Actionability	Provides copy-paste ready commands (gh pr view/diff with exact JSON fields), a fully specified three-step L0–L3 leveling algorithm, a fixed four-question remediation checklist with evidence requirements, and a verbatim output template. Common cases are covered by concrete rules rather than abstract direction.	5 / 5
Workflow Clarity	A 7-step workflow sequences three explicitly named passes, each cross-referenced to its own section, with explicit validation checkpoints (before/after file:line evidence per row, independent-list-vs-author-table comparison, demotion rules for insufficient evidence). The skill is read-only review, so the destructive-operation cap does not apply.	5 / 5
Progressive Disclosure	Sections are well organized and everything inlined is needed each run, but the skill is a single ~185-line monolith with the 模板逐节核对 detail that could live in a separate reference, and the one external file (surfaces.md) is referenced three times with a clear role yet no stated path or location. Good structure with minor organization gaps, not the well-signaled one-level-deep reference layout of anchor 5.	4 / 5
	Total	19 / 20 Passed

Description

83%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A well-scoped, concrete description that clearly states what the skill does and exactly when to use it, with an explicit repo-scoped trigger. Its main weakness is thin trigger-term variety (no synonyms for 'review PR') and mild overlap with generic code-review triggers.

Suggestions

Add one or two natural trigger variations (e.g., 'code review' / '评审 PR' / '审查这个 PR') so the description matches more of the phrasings a user would naturally say.

Briefly mention the three-pass structure or the L0–L3 impact-level output in the description so the what-clause covers the skill's full scope, not just its headline actions.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions — "按 PR 模板逐节核对描述与 diff", "单独做一轮'回归与用户影响'评审", "给出影响等级与评审摘要", plus a summary-language rule — but stops short of the full process (three passes, remediation checks, output format), so it falls at 'several specific actions with minor gaps' rather than comprehensive coverage.	4 / 5
Completeness	Explicitly answers both: "在 kimi-code 仓库里 review 一个 PR 时使用" is a concrete when-clause, and the what-clause enumerates four specific capabilities. It matches the anchor-5 example structure (what + 'use when' with concrete triggers), not the weaker 'when could be more explicit' of anchor 4.	5 / 5
Trigger Term Quality	Natural phrases a user would say are present ("review 一个 PR", "kimi-code 仓库"), giving good keyword coverage, but common variations like "code review", "评审", or "diff 审查" are missing, keeping it below the comprehensive-synonyms anchor.	4 / 5
Distinctiveness Conflict Risk	Scoping to the kimi-code repo plus the distinctive '回归与用户影响' framing gives it a clear niche, but the generic "review 一个 PR" trigger overlaps with general PR-review/code-review skills — minor overlap with closely related skills rather than minimal conflict risk.	4 / 5
	Total	17 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: MoonshotAI/kimi-code
Path: .agents/skills/review-pr/SKILL.md
Commit: 21406fb

Reviewed: 2 days ago

Table of Contents

Discovery Implementation Validation

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.