experiment-audit

Audit experiment integrity before claiming results. Uses cross-model review (external reviewer backend) to check for fake ground truth, score normalization fraud, phantom results, and insufficient scope. Use when user says "审计实验", "check experiment integrity", "audit results", "实验诚实度", or after experiments complete before writing claims.

Quality

88%

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

81%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 content is a well-structured, actionable integrity-audit workflow with explicit validation checkpoints and clear cross-references. Its main weakness is moderate verbosity from the inlined full reviewer prompt and repeated backend-calling conventions.

Suggestions

Move the ~70-line verbatim reviewer prompt into a referenced file (e.g. references/auditor-prompt.md) and inline only the variable bindings, trimming conciseness cost.

Consolidate the duplicated codex/manual calling conventions so each backend's parameters appear once rather than restated across 'Reviewer Calling Convention' and Step 2.

Add a references/ directory to hold reviewer-routing, review-tracing, and prompt material so progressive disclosure can offload detail that is currently inlined.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and assumes competence—no padding about what 'integrity' means—but the full verbatim reviewer prompt and duplicated backend-calling conventions add length that could be trimmed via a shared reference, so it is not fully lean.	4 / 5
Actionability	Provides concrete MCP call signatures with exact config keys, a copy-paste reviewer prompt, and explicit output templates for both .md and .json; minor gaps are placeholders like '[today]' and '[project name]' that require manual substitution.	4 / 5
Workflow Clarity	The four-step workflow is explicitly sequenced (collect paths → send to reviewer → parse/write report → print summary) with PASS/WARN/FAIL verdict checkpoints and clear validation of reviewer identity, satisfying the audit checklist and error-handling guidance.	5 / 5
Progressive Disclosure	Good section structure with one-level-deep references to shared-references/*.md for cadence, reviewer independence, and tracing; the large inline reviewer prompt could arguably live in a separate file, and no bundle files exist to offload detail.	4 / 5
	Total	17 / 20 Passed

Description

95%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 states what the skill does, lists concrete fraud checks, and gives explicit bilingual and behavioral triggers. It is concise, third-person, and clearly distinct from other skills.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions—'check for fake ground truth, score normalization fraud, phantom results, and insufficient scope'—with named fraud patterns rather than vague language, though it describes one composite audit task rather than a broad action set.	4 / 5
Completeness	Explicitly answers both 'what' (audit experiment integrity via cross-model review for four named fraud patterns) and 'when' (an explicit 'Use when user says... or after experiments complete before writing claims' clause with concrete trigger phrases).	5 / 5
Trigger Term Quality	Comprehensive natural trigger coverage including exact quoted phrases ('审计实验', 'check experiment integrity', 'audit results', '实验诚实度') plus a behavioral trigger ('after experiments complete before writing claims'), spanning synonyms and bilingual variants.	5 / 5
Distinctiveness Conflict Risk	Clear niche—experiment-integrity auditing via cross-model review—with distinctive bilingual trigger phrases unlikely to collide with other skills; the 'experiment audit' framing is highly specific.	5 / 5
	Total	19 / 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/experiment-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.