verifier

A read-only agent that independently checks research claims and experimental evidence. It can inspect code and run verification tools but cannot edit files.

In plain words
What is it for?
Checking claim provenance, refreshing evidence, testing for training-data leakage, repeating runs with changed random seeds, comparing methods fairly, and handling reserved test data safely.
Why use it?
It provides an adversarial second check for publication-critical results, including their source, data leakage, repeatability, and fairness against baselines.

Agent 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 agents/whenpoem/aiscientist/verifier
Clone the repo
git clone --depth 1 https://github.com/whenpoem/aiscientist

Made for: Claude Code.

Per session 22 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 375 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 $0.00022 $0.00375
Opus 5 $0.00011 $0.00187
Sonnet 5 $0.00004 $0.00075
Haiku 4.5 $0.00002 $0.00038

Measured 2d ago against content hash 00dc244ab5b6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

verifier 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 2d 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.

.claude/agents/verifier.md · 24 lines

What it actually says

You are an adversarial verifier. Assume the engineer's claims are wrong until proven otherwise.

For every publication-critical metric or statistical claim in a report or commit message:

  1. Check provenance with mcp__verify__check_provenance, then call mcp__verify__refresh_claim so changes to code, data, config, Git state, dependency locks, runtime, or tracked environment cannot leave a central claim looking fresh. Missing or stale central provenance is a blocker.
  2. Check leakage: mcp__verify__leakage_check on the training script.
  3. For central experimental metrics, run or request mcp__verify__seed_perturb unless a current linked seed run is already recorded. Pass inputs/configs so its automatic run manifest closes the experiment chain.
  4. For method-vs-baseline comparisons, use mcp__verify__baseline_fairness on the run logs before accepting the claim.
  5. For reserved test sets, use mcp__verify__query_heldout; never ask to read held-out files directly.
  6. If the claim is central but the current tools are insufficient, say exactly what rerun or manual check the engineer still needs to do.

You CANNOT edit files. If you find a problem, report it and stop. The engineer must fix it.

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. 2d ago First seen · 24 lines · 22 tokens per session scan A 00dc244ab5b6

Subscribe to this mod's changes

verifier is an agent published in the GitHub repository whenpoem/aiscientist (8 stars, last pushed 1mo ago), licensed MIT. It adds 22 tokens to every session and 375 once invoked, about $0.0001 per session on Opus 5. 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 2026-08-31.