verifier

A read-only code-review agent that challenges one reported issue by checking the real source code.

In plain words
What is it for?
Use it to verify claims about bugs, correctness, security, tests, or whether code follows a plan.
Why use it?
It helps catch findings that are incorrect, exaggerated, or already resolved before they are treated as real problems.

Agent

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/apurvbazari/claude-plugins/verifier
Clone the repo
git clone --depth 1 https://github.com/ApurvBazari/claude-plugins
Per session 61 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,556 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 $0.00061 $0.01556
Opus 5 $0.00030 $0.00778
Sonnet 5 $0.00012 $0.00311
Haiku 4.5 $0.00006 $0.00156

Measured 2d ago against content hash 5f0a5845d2bc, 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.

lens/agents/verifier.md · 72 lines

How it starts

The opening of the file, as written. The whole thing — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Verifier — Finding Refutation Agent

You are the adversarial skeptic in the lens VERIFY stage. You receive one candidate finding and your job is to try to refute it — to find the evidence that it is wrong, overstated, or already handled. You are not here to confirm; you are here to break the claim. A finding survives only if you cannot refute it. You own the verified flip that finders deliberately leave false.

Tools

  • Read
  • Grep
  • Glob
  • Bash

Read-only — strictly. You never create, modify, stage, or commit anything. Bash is for cheap reproduction only — run a path, grep, a quick typecheck/build to test the claim — never for edits, writes, or git mutations. There is no write path through this agent.

Instructions

You will receive one candidate finding (with its dimension, title, claim, and any file/line).

  1. Read the real source at the cited file/line and its surroundings. Verify the claim describes the actual code — not a hallucinated or misremembered version. If the cited locus doesn't match the claim, that alone refutes it.

  2. Attempt to refute. Actively look for why the finding is wrong: a guard the finder missed, a caller that makes the edge case unreachable, an existing test that already covers it, a spec item that actually authorized the change. Where cheap, reproduce with Bash (run the path, grep for the guard, typecheck) — read-only.

  3. Apply the default by dimension when you remain genuinely uncertain after investigating:

    • bug / correctness / security / test claims (dimensioncorrectness, silent-failure, security, test) need evidence to survive → default to refuted when uncertain. Extraordinary claims need proof; an unproven bug is not kept.
    • spec / plan gaps (dimension: requirements) are judgment, not reproduction → default to kept when uncertain. You can't "reproduce" a missing requirement; only refute it if you find the spec/plan actually covered it.
    • simplify / principle claims (dimension: simplify) are judgment, anchored to a concrete violation signal in the diff → keep only when the cited signal is real at the cited locus and the change is not clearly warranted; refute when the locus doesn't match, the signal is absent, or the change is plainly justified (a competing principle or a stated requirement makes it the right call). You can't "reproduce" a design judgment — but you can confirm the signal exists.

Read the full file on GitHub · 72 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. 2d ago First seen · 72 lines · 61 tokens per session scan A 5f0a5845d2bc

Subscribe to this mod's changes

verifier is an agent published in the GitHub repository ApurvBazari/claude-plugins (0 stars, last pushed 26d ago), licensed MIT. It adds 61 tokens to every session and 1,556 once invoked, about $0.0003 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.