citation-audit

Zero-context verification that every bibliographic entry in the paper is real, correctly attributed, and used in a context the cited paper actually supports — catching hallucinated authors, wrong years, fabricated venues, version mismatches, and wrong-context citations. Use when user says "审查引用", "check citations", "citation audit", "verify references", "引用核对", or before submission to ensure bibliography integrity.

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Advisory

Suggest reviewing before use

Quality

Content

77%

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

The body is highly actionable with a clear, validated multi-step workflow and concrete code/schemas, but it is verbose with repetition (soft-only and uncited modes restated multiple times) and monolithic — skill-specific detail that could live in separate reference files is inlined rather than progressively disclosed.

Suggestions

Consolidate the soft-only mode material into one section; it is currently restated in Constants, Key Rules, the dedicated 'Optional: Soft-Only Mode' section, and the augmented JSON schema — reference one authoritative location from the others to cut repetition (improves conciseness).

Move the heavy detail blocks (the augmented soft-only JSON schema, the full Submission Artifact Emission schema, and the Uncited Entry Detection protocol) into bundled reference files under references/ and link to them one level deep from SKILL.md, turning the file into a concise overview (improves progressive_disclosure).

Extract the verdict translation table and decision table into a single reference doc and summarize inline, reducing the inline schema/boilerplate that drives the file's length (improves conciseness).

Dimension	Reasoning	Score
Conciseness	The body is ~500 lines / 30KB with genuine operational density and no concept-explanation fluff (it does not explain what a citation is), so it is above score 1; however soft-only mode is restated in four places (Constants, Key Rules, a dedicated section, and the augmented schema) and uncited detection is repeated, which is tighten-able verbosity per the score-2 anchor rather than the lean score-3 anchor.	2 / 3
Actionability	It provides a complete, copy-paste-ready mcp__codex__codex invocation block (model, config, sandbox, full prompt), concrete JSON ledger schemas with worked examples, and executable `latexmk -C && latexmk -pdf -interaction=nonstopmode main.tex` commands, matching the score-3 fully-executable anchor.	3 / 3
Workflow Clarity	The seven-step workflow is clearly sequenced with an explicit validation checkpoint in Step 7 (recompile and confirm 'No new Citation undefined warnings', 'No Reference undefined warnings') plus a verdict decision table, matching the score-3 anchor with explicit validation steps and error-recovery feedback rather than the gap-prone score-2 anchor.	3 / 3
Progressive Disclosure	It is a monolithic single file with skill-specific heavy detail kept inline (soft-only mode ~80 lines, uncited detection ~35, submission-artifact emission ~90) that could be split into bundled reference files; the only references point to external ../shared-references/*.md docs that are not part of this skill's bundle and cannot be verified, so it sits at the score-2 'content that should be separate is inline' anchor rather than the appropriately-split score-3 anchor.	2 / 3
	Total	10 / 12 Passed

Description

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.

The description is strong: third-person voice, specific capabilities, natural bilingual trigger terms, and an explicit 'Use when…' clause covering both what and when. It is clearly distinguishable from sibling audit skills via citation-specific triggers.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete verification actions ('every bibliographic entry… is real, correctly attributed, and used in a context the cited paper actually supports') and enumerates specific problems caught ('hallucinated authors, wrong years, fabricated venues, version mismatches, and wrong-context citations'), matching the score-3 anchor for multiple specific concrete actions rather than the score-2 single-domain anchor.	3 / 3
Completeness	It explicitly answers both 'what' (zero-context verification of bibliographic entries) and 'when' via an explicit 'Use when user says…' trigger clause, satisfying the score-3 anchor and not the score-2 cap which applies when 'when' is missing or only implied.	3 / 3
Trigger Term Quality	It provides natural trigger phrases users would actually say ('check citations', 'citation audit', 'verify references', plus bilingual '审查引用'/'引用核对'), giving good coverage of common variations per the score-3 anchor rather than the partial score-2 anchor.	3 / 3
Distinctiveness Conflict Risk	It carves a clear bibliographic-integrity niche with citation-specific triggers ('check citations', 'verify references'), making it unlikely to fire for unrelated skills per the score-3 anchor rather than the overlapping score-2 anchor.	3 / 3
	Total	12 / 12 Passed

Validation

75%

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 — 12 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (503 lines); consider splitting into references/ and linking	Warning
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 1 suspicious	Warning

	Total	12 / 16 Passed

Repository: wanshuiyin/Auto-claude-code-research-in-sleep
Commit: 82076e5

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.