experiment-audit

experiment-audit is a skill for Claude Code from zjunlp/Mechanist. It costs 115 tokens per session (4,347 once invoked), scanned A, original, MIT.

A review process for checking whether one experimental claim is supported by a sound method. It checks the source of the reference answers, score calculations, result files, unused code, evaluation scope, and evaluation type, then returns PASS, WARN, or FAIL.

In plain words
What is it for?
Auditing one research or machine-learning claim before reporting it. An external language-model reviewer examines the relevant code and files independently.
Why use it?
It helps catch common problems such as made-up reference answers, misleading score normalization, missing results, and claims based on too little data.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: reads .claude/ paths.

Part of the mechanist plugin — 54 skills, 4 agents shipped together

Good fit Auditing one research or machine-learning claim before reporting it. An external language-model reviewer examines the relevant code and files independently.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/zjunlp/mechanist/experiment-audit
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add zjunlp/Mechanist --skill experiment-audit
Clone the repo
git clone --depth 1 https://github.com/zjunlp/Mechanist

Made for: Claude Code.

Or install mechanist, the plugin that ships this one along with the rest of its 54 skills, 4 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for experiment-audit

README.md
[![agentmods](https://agentmods.dev/badge/skills/zjunlp/mechanist/experiment-audit.svg)](https://agentmods.dev/skills/zjunlp/mechanist/experiment-audit)
Your own site
<a href="https://agentmods.dev/skills/zjunlp/mechanist/experiment-audit"><img src="https://agentmods.dev/badge/skills/zjunlp/mechanist/experiment-audit.svg" alt="Measured on agentmods" height="20"></a>
Per session 115 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,347 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 2 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 9 findings, up to high

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • high Agent Snooping · line 53
    Skill accesses MCP server configuration files (mcp.json). MCP configs contain server URLs, authentication tokens, and tool definitions — reading them allows the skill to discover and potentially abuse other tool integrations.
    Fix: Remove all code or instructions that read MCP configuration files (mcp.json). MCP server details should be managed by the agent runtime, not read by individual skills.
  • high Agent Snooping · line 54
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • high Agent Snooping · line 66
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • high Agent Snooping · line 67
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • high Agent Snooping · line 68
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • high Agent Snooping · line 69
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • high Agent Snooping · line 70
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • high Agent Snooping · line 78
    Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
    Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
  • medium Rogue Agent · line 36
    Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
    Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00115 $0.04347
Opus 5 $0.00057 $0.02174
Sonnet 5 $0.00023 $0.00869
Haiku 4.5 $0.00012 $0.00435

Measured 8d ago against content hash 85640ecd9cf2, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

experiment-audit scanned grade A with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Reads agent configuration directorieslowAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

2. **User MCP config** — `~/.claude/settings.json`, same field.

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

Reads MCP configurationlowAgent snooping

mcp.json carries server URLs and auth tokens; reading it lets a mod discover and abuse other integrations.

1. **Project MCP config** — `${PROJECT_ROOT}/.mcp.json`, field `mcpServers["llm-chat"].env.{LLM_MODEL,LLM_BASE_URL,LLM_API_KEY}`.

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

skills/experiment-audit/SKILL.md · 354 lines

How it starts

The opening of the file, as written. The whole thing — 354 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Experiment Audit: Per-Claim Cross-Model Integrity Verification

Audit the experimental process for one claim: $ARGUMENTS

Why This Exists

LLM agents can produce fraudulent experimental results through:

  1. Fake ground truth — creating synthetic "reference" from model outputs, then reporting high agreement as performance
  2. Score normalization — dividing metrics by the model's own max to get 0.99+
  3. Phantom results — claiming numbers from files that don't exist or functions never called
  4. Insufficient scope — reporting 2-scene pilots as "comprehensive evaluation"

These are NOT intentional deception — they are failure modes of optimizing agents that lack integrity constraints. This skill adds that constraint, once per claim: each invocation scopes to one claim's runs and returns a PASS/WARN/FAIL verdict on that claim's experimental process.

Core Principle

The executor (Claude) collects file paths scoped to the target claim. The external LLM reviewer reads code and judges integrity. The executor does NOT participate in integrity judgment.

This follows shared-references/reviewer-independence.md and shared-references/experiment-integrity.md.

Constants

  • REVIEWER_BACKEND = llm-chat — External LLM reviewer via llm-chat MCP (model defers to LLM_MODEL env). Always ask the external reviewer for strict, high-rigor feedback. Override with — reviewer: oracle-pro via Oracle MCP.

Arguments

  • <experiment-dir-or-results-path> (positional, required) — directory containing the experiment artifacts to audit. Typically refine-logs/ (for main-experiment audit) or verify/<claim_dir>/variants/ (for variant audit).
  • — claim <Cx> (required) — the claim whose experimental process is being audited. The skill is inherently per-claim: each invocation produces ONE PASS/WARN/FAIL verdict for the named claim, derived from auditing only that claim's linked milestones/runs. To audit N claims, call the skill N times.
  • — output-dir <path> (optional) — directory to write EXPERIMENT_AUDIT.{md,json} into. Defaults to current working directory. When given, the skill creates <path> if needed and writes directly to <path>/EXPERIMENT_AUDIT.md and <path>/EXPERIMENT_AUDIT.json; callers do not need a follow-up mv.

Read the full file on GitHub · 354 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 8d ago First seen · 354 lines · 115 tokens per session scan A 85640ecd9cf2

Subscribe to this mod's changes

experiment-audit is a skill published in the GitHub repository zjunlp/Mechanist (72 stars, last pushed 12d ago), licensed MIT. It adds 115 tokens to every session and 4,347 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 2 findings (reads agent configuration directories, reads mcp configuration). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.