audit-candidate

audit-candidate is a command for Claude Code from ProjectDXAI/labrat. It costs 0 tokens per session (167 once invoked), scanned A, original, MIT.

A command for reviewing a suspicious software experiment candidate to determine whether it contains a real bug, regression or incorrect classification.

In plain words
What is it for?
It helps inspect the highest-scoring suspicious candidate, write an audit note, and return the candidate to the search queue when the evidence supports doing so.
Why use it?
It replaces an informal judgment with an audit based on the candidate’s recorded results, implementation notes and checkpoints. It also keeps authoritative scores unchanged.

Command for Claude Code

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.

agentmods
npx agentmods add commands/projectdxai/labrat/audit-candidate
Clone the repo
git clone --depth 1 https://github.com/ProjectDXAI/labrat

Made for: Claude Code.

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 audit-candidate

README.md
[![agentmods](https://agentmods.dev/badge/commands/projectdxai/labrat/audit-candidate.svg)](https://agentmods.dev/commands/projectdxai/labrat/audit-candidate)
Your own site
<a href="https://agentmods.dev/commands/projectdxai/labrat/audit-candidate"><img src="https://agentmods.dev/badge/commands/projectdxai/labrat/audit-candidate.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 167 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00000 $0.00167
Opus 5 $0.00000 $0.00084
Sonnet 5 $0.00000 $0.00033
Haiku 4.5 $0.00000 $0.00017

Measured 4d ago against content hash 86bb67ed7103, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

audit-candidate scanned grade A with 0 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 4d 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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

examples/nlp-sentiment/research_lab/.claude/commands/audit-candidate.md · 10 lines

What it actually says

Walk the highest-signal suspicious candidate through an implementation audit.

  1. Read state/frontier.json.audit_queue and state/frontier.json.invalid_fast_candidates.
  2. Pick the candidate with the highest search_eval among the suspicious set. Read its record from state/candidates.jsonl and its evaluation from state/evaluations.jsonl, plus any checkpoints.jsonl under its artifact_dir.
  3. Read implementation_audit.md.
  4. Decide: is this a real bug, a real regression, or a misclassification?
  5. Leave an audit note under logs/audits/<candidate_id>.md and, if the candidate should re-enter the frontier, clear it from audit_queue with the appropriate runtime command.

Do not change scores. The runtime is authoritative.

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. 4d ago First seen · 10 lines · 0 tokens per session scan A 86bb67ed7103

Subscribe to this mod's changes

audit-candidate is a command published in the GitHub repository ProjectDXAI/labrat (238 stars, last pushed 26d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 167 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.