research-refine

Turn a vague research direction into a problem-anchored, elegant, frontier-aware, implementation-oriented method plan via iterative Claude review. Use when the user says "refine my approach", "帮我细化方案", "decompose this problem", "打磨idea", "refine research plan", "细化研究方案", or wants a concrete research method that stays simple, focused, and top-venue ready instead of a vague or overbuilt idea.

Quality

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Quality

Content

77%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A thorough, highly actionable iterative-refinement workflow with excellent sequencing and validation checkpoints. Its main weaknesses are redundancy between Overview and Key Rules and a monolithic structure that inlines large templates instead of splitting them into reference files.

Suggestions

Deduplicate the four principles: keep them in Overview and have 'Key Rules' reference them instead of restating each one verbatim.

Move the large proposal markdown template and the Phase 2 reviewer prompt into separate reference files (e.g. references/proposal-template.md, references/reviewer-prompt.md) and link to them, reducing SKILL.md to an overview.

Trim rhetorical padding such as 'The goal is not to produce a bloated proposal or a benchmark shopping list' that does not change the instruction.

Dimension	Reasoning	Score
Conciseness	Mostly efficient for a complex multi-phase skill, but the four Overview principles are restated almost verbatim in the 'Key Rules' section, and rhetorical padding ('The goal is not to produce a bloated proposal or a benchmark shopping list') plus repeated 'Composing with Other Skills' context add tokens that could be trimmed.	3 / 5
Actionability	Fully executable: concrete MCP tool calls with full syntax (mcp__claude-review__review_start / review_reply_start / review_status), complete copy-paste-ready reviewer prompts, JSON checkpoint schemas, and detailed markdown output templates with real file paths.	5 / 5
Workflow Clarity	Phases 0-5 are clearly sequenced with an explicit checkpoint (REFINE_STATE.json) after each phase, a stop condition (overall score >= SCORE_THRESHOLD or MAX_ROUNDS), and a validate -> fix -> re-review feedback loop with checkpoint-recovery logic on resume.	5 / 5
Progressive Disclosure	Good section structure and clearly signaled shared-references links, but the skill is monolithic at ~730 lines with the large proposal template (~90 lines) and reviewer prompt (~70 lines) inlined rather than split into reference files; no bundle files are provided.	3 / 5
	Total	16 / 20 Passed

Description

86%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 strong description with explicit what/when structure and rich bilingual trigger phrases. It is slightly over-adjectivized on the 'what' side, listing qualities of the output rather than multiple discrete actions.

Dimension	Reasoning	Score
Specificity	Names the domain (research method planning) and one concrete action ('Turn a vague research direction into a ... method plan via iterative Claude review'), but the remaining adjectives (problem-anchored, elegant, frontier-aware) are descriptors rather than distinct concrete actions, so coverage is not comprehensive.	3 / 5
Completeness	Explicitly answers both what ('Turn a vague research direction into a ... method plan via iterative Claude review') and when ('Use when the user says ... or wants a concrete research method ...'), with concrete trigger phrases.	5 / 5
Trigger Term Quality	Comprehensive natural trigger phrases including synonyms and bilingual variants ('refine my approach', 'decompose this problem', 'refine research plan', '帮我细化方案', '打磨idea', '细化研究方案') that users would actually say.	5 / 5
Distinctiveness Conflict Risk	Clear niche (iterative research-method refinement) with distinct 'refine' triggers, but minor overlap risk with sibling research skills (/idea-creator, /experiment-plan) and the somewhat generic 'decompose this problem'.	4 / 5
	Total	17 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (733 lines); consider splitting into references/ and linking	Warning
relative_links	Relative link issues: 4 suspicious	Warning

	Total	14 / 16 Passed

Repository: wanshuiyin/Auto-claude-code-research-in-sleep
Path: skills/skills-codex-claude-review/research-refine/SKILL.md
Commit: f4f20f9

Reviewed: 1 day 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.