paper-claim-audit

Zero-context verification that every number, comparison, and scope claim in the paper matches raw result files. Uses a fresh cross-model reviewer with NO prior context to prevent confirmation bias. Use when user says "审查论文数据", "check paper claims", "verify numbers", "论文数字核对", or before submission to ensure paper-to-evidence fidelity.

Quality

75%

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

Fix and improve this skill with Tessl

tessl review fix ./skills/paper-claim-audit/SKILL.md

Quality

Content

67%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, actionable audit workflow with a strong reviewer prompt and clear sequencing, weakened mainly by repetition of the thread-independence invariant and a heavy inlined contract schema. Consolidating the repeated invariants and offloading the JSON schema detail would lift conciseness.

Suggestions

State the 'fresh thread, never codex-reply, no prior context' invariant once in Key Rules and reference it from Step 2 / Thread independence instead of restating it three times.

Move the full assurance-contract JSON schema and path-convention detail into shared-references/assurance-contract.md (already referenced) and keep only the verdict decision table inline in SKILL.md.

Add an explicit internal checkpoint (e.g., 'if reviewer returns malformed output, re-invoke on a fresh thread with a tightened prompt') so the workflow has a visible validate→retry loop.

Dimension	Reasoning	Score
Conciseness	The 'fresh thread / never codex-reply' invariant is restated in Step 2, Key Rules, and a dedicated Thread independence subsection, and the full assurance-contract JSON schema is inlined, making it mostly efficient but with noticeable repetition that could be tightened; not as padded as the 2 anchor but not lean enough for 4.	3 / 5
Actionability	Provides a copy-paste Codex prompt with a full audit protocol, seven enumerated failure modes, an output format, plus report and summary templates and an explicit /render-html command — mostly executable with minor gaps from bracketed placeholders and deferrals to shared-references files.	4 / 5
Workflow Clarity	Four clearly numbered steps (Collect Files → Fresh Reviewer Audit → Write Report → Print Summary) are reinforced by a verdict decision table and a failure-mode checklist; the skill is read-only so the destructive-cap does not apply, but there is no internal validate→fix→retry loop to reach 5.	4 / 5
Progressive Disclosure	Well-organized sections with clearly signaled one-level-deep references to shared-references/*.md (external-cadence, review-tracing, assurance-contract, reviewer-independence, integration-contract); the inlined JSON schema and path conventions are slightly heavy but appropriately authoritative, so it sits above the 3 anchor and below the ideal 5.	4 / 5
	Total	15 / 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 strong, well-targeted description that clearly states what the skill does and when to invoke it, with natural bilingual trigger phrases. Minor room to broaden trigger synonyms and sharpen action specificity.

Dimension	Reasoning	Score
Specificity	Names the domain and several concrete actions — 'verification that every number, comparison, and scope claim in the paper matches raw result files' and 'fresh cross-model reviewer with NO prior context' — but stops short of fully enumerating every action, so it sits below the comprehensive 5 anchor.	4 / 5
Completeness	Explicitly answers both what (zero-context verification of every claim against raw result files via a fresh cross-model reviewer) and when (concrete trigger phrases plus 'before submission'), matching the clear-what-and-when-with-triggers anchor.	5 / 5
Trigger Term Quality	Includes several natural bilingual trigger phrases ('审查论文数据', 'check paper claims', 'verify numbers', '论文数字核对', 'before submission'), giving good keyword coverage, though it lacks broader synonyms/file-extension variants that would reach the 5 anchor.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (paper-to-evidence fidelity) with distinct triggers, but overlaps slightly with sibling audit skills (experiment-audit, result-to-claim), keeping it just below the minimal-conflict 5 anchor.	4 / 5
	Total	17 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
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	13 / 16 Passed

Repository: wanshuiyin/Auto-claude-code-research-in-sleep
Path: skills/paper-claim-audit/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.