auto-paper-improvement-loop

Autonomously improve a generated paper via Claude review through claude-review MCP → implement fixes → recompile, for 2 rounds. Use when user says "改论文", "improve paper", "论文润色循环", "auto improve", or wants to iteratively polish a generated paper.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

70%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.

The body is a well-sequenced, highly actionable workflow with strong validation checkpoints, but it is held back by verbosity (a duplicated review prompt) and weak progressive disclosure (no bundle files; large reference-style content inlined). Actionability is strong with only minor executable gaps.

Suggestions

De-duplicate the reviewer prompt: extract the shared prompt template into a reference file (e.g. references/review-prompt.md) and have Steps 2 and 5 reference it with their round-specific deltas, removing ~45 lines of repetition.

Move the edit-whitelist specification (schema, resolution rules, glob semantics, detectors) into a separate references/edit-whitelist.md and keep only a concise summary plus a pointer in SKILL.md, improving progressive disclosure.

Complete the Step 4.5 normalization script by implementing the theorem-block comparison logic instead of leaving it as a comment stub, so the regression check is copy-paste executable.

Dimension	Reasoning	Score
Conciseness	Mostly high-signal domain guidance, but the ~45-line reviewer prompt is duplicated verbatim between Step 2 and Step 5, and several "Empirical motivation" paragraphs are padded, so it could be tightened rather than being lean throughout.	3 / 5
Actionability	Provides concrete latexmk/pdfinfo/grep commands, full MCP call templates, and fix-pattern tables, but the Step 4.5 Python snippet defines normalize() and leaves the actual comparison logic as a comment stub, and review prompts carry unfilled [VENUE]/[list files] placeholders.	4 / 5
Workflow Clarity	Clear Step 0 through Step 9 sequence with explicit validation checkpoints (recompile verification, restatement regression test, format-check stop criteria, edit-whitelist gate) and feedback loops for error recovery, satisfying the destructive/batch validation requirement.	5 / 5
Progressive Disclosure	Good section headers, but no bundle files exist and the skill is a monolithic ~620 lines with reference-style content inlined (the ~100-line whitelist spec and duplicated review prompts) that would be better split into separate reference files.	3 / 5
	Total	15 / 20 Passed

Description

96%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.

The description is strong: it concisely states the loop's concrete actions, gives explicit bilingual trigger phrases, and clearly answers both what and when. The only soft spot is minor distinctiveness risk from the generic "auto improve" trigger and overlap with the related /auto-review-loop skill.

Dimension	Reasoning	Score
Specificity	Names the domain and multiple concrete actions ("Claude review through claude-review MCP", "implement fixes", "recompile, for 2 rounds"), giving comprehensive coverage of the loop's behavior rather than just several actions with gaps.	5 / 5
Completeness	Explicitly states both what it does (review→fix→recompile for 2 rounds) and when to use it ("Use when user says ..."), with concrete trigger phrases, matching the anchor for clearly answering both what and when.	5 / 5
Trigger Term Quality	Provides several natural trigger phrases across English and Chinese including synonyms ("improve paper", "auto improve", "改论文", "论文润色循环", "iteratively polish a generated paper"), matching comprehensive natural-term coverage.	5 / 5
Distinctiveness Conflict Risk	Targets a distinct paper-improvement niche with paper-specific triggers, but "auto improve" is generic and the body itself notes overlap with the adjacent /auto-review-loop skill, indicating minor overlap risk with a closely related skill.	4 / 5
	Total	19 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (626 lines); consider splitting into references/ and linking	Warning

	Total	15 / 16 Passed

Repository: wanshuiyin/Auto-claude-code-research-in-sleep
Path: skills/skills-codex-claude-review/auto-paper-improvement-loop/SKILL.md
Commit: 0472e53

Reviewed: 5 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.