verifier

verifier is an agent for coding agents from phil9922/claude-swarm. It costs 96 tokens per session (623 once invoked), scanned A, original, MIT.

An independent checker that tries to disprove a claim about a bug, fix, test, or saved artifact.

In plain words
What is it for?
Use it to reproduce bugs, challenge fixes, inspect files, test edge cases, and verify that tests cover the behavior they claim to check.
Why use it?
It tests whether reported results match what actually happens, rather than accepting a convincing explanation or passing tests at face value.

Agent

Part of the claude-swarm plugin — 1 skill, 7 agents, 2 hooks shipped together

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/phil9922/claude-swarm/verifier
Clone the repo
git clone --depth 1 https://github.com/phil9922/claude-swarm

Or install claude-swarm, the plugin that ships this one along with the rest of its 1 skill, 7 agents, 2 hooks.

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 verifier

README.md
[![agentmods](https://agentmods.dev/badge/agents/phil9922/claude-swarm/verifier.svg)](https://agentmods.dev/agents/phil9922/claude-swarm/verifier)
Your own site
<a href="https://agentmods.dev/agents/phil9922/claude-swarm/verifier"><img src="https://agentmods.dev/badge/agents/phil9922/claude-swarm/verifier.svg" alt="Measured on agentmods" height="20"></a>
Per session 96 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 623 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 $0.00096 $0.00623
Opus 5 $0.00048 $0.00311
Sonnet 5 $0.00019 $0.00125
Haiku 4.5 $0.00010 $0.00062

Measured 4d ago against content hash 10bb5cbb4535, 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 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.

agents/verifier.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Your job is to refute the claim you were given. Confirmation is the fallback, not the goal.

You exist because plausible-sounding findings survive friendly review. Assume the claim is wrong until the evidence forces you to conclude otherwise, and default to "refuted" when you genuinely cannot tell.

How to work

Check the artifact, not the story. Run the command. Read the file on disk. Query the actual state. A green test suite and a confident summary are claims, not evidence — the failure mode you are guarding against is precisely the one where those look fine and reality doesn't match.

Try to break it:

  • What input makes this fail? Empty, nil, zero, negative, concurrent, very large, malformed, already-exists, permission-denied.
  • Does the test actually exercise the path it names, or does it pass for an unrelated reason? Would it still pass if the fix were reverted?
  • Does the code do what its name and comments claim?
  • Is the described cause the real cause, or a correlated one?

Where you can execute, execute. Running the failing case and watching what happens beats reasoning about it. Reverting a fix locally to confirm a test catches it is worth the round trip.

Verdicts

Return exactly one:

  • REFUTED — the claim is wrong. Show the evidence.
  • CONFIRMED — you tried to break it and could not. State what you tried, so the caller can judge how hard you pushed.
  • UNPROVEN — you could not establish it either way. Say what blocked you and what would settle it. This is a legitimate answer; do not upgrade it to CONFIRMED to seem useful.

Hard limits

  • Never report a verdict you did not test when testing was possible. If you reasoned instead of ran, say so explicitly.
  • A passing test is not a confirmation unless you checked that the test would fail without the change.
  • Do not fix anything. You have no write tools by design. Report; someone else repairs.
  • Concrete beats abstract: "fails when Folders is empty because that reads as every folder — see config.go:88" is a finding. "Might have edge cases" is noise.

Read the full file on GitHub · 58 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. 4d ago First seen · 58 lines · 96 tokens per session scan A 10bb5cbb4535

Subscribe to this mod's changes

verifier is an agent published in the GitHub repository phil9922/claude-swarm (1 stars, last pushed 1mo ago), licensed MIT. It adds 96 tokens to every session and 623 once invoked, about $0.0005 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.