claim-verifier

claim-verifier is an agent for coding agents from kbichave/skills. It costs 91 tokens per session (972 once invoked), scanned A, original, MIT.

A code-review agent that checks claims in a review panel’s merged findings against current official documentation on the web. It covers claims about frameworks, libraries, APIs, statistics, SQL dialects, versions, and security classifications.

In plain words
What is it for?
Use it after other reviewers produce findings that need confirmation. It is especially relevant when a finding depends on current library behavior, a database engine’s SQL rules, or a CWE or OWASP mapping.
Why use it?
It reduces the risk of reporting a bug based on outdated, incorrect, or platform-specific assumptions. It is the panel’s only network-based verification stage.

Agent

Part of the deep plugin — 4 skills, 17 agents, 6 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/kbichave/skills/claim-verifier
Clone the repo
git clone --depth 1 https://github.com/kbichave/skills

Or install deep, the plugin that ships this one along with the rest of its 4 skills, 17 agents, 6 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 claim-verifier

README.md
[![agentmods](https://agentmods.dev/badge/agents/kbichave/skills/claim-verifier.svg)](https://agentmods.dev/agents/kbichave/skills/claim-verifier)
Your own site
<a href="https://agentmods.dev/agents/kbichave/skills/claim-verifier"><img src="https://agentmods.dev/badge/agents/kbichave/skills/claim-verifier.svg" alt="Measured on agentmods" height="20"></a>
Per session 91 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 972 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 $0.00091 $0.00972
Opus 5 $0.00046 $0.00486
Sonnet 5 $0.00018 $0.00194
Haiku 4.5 $0.00009 $0.00097

Measured today against content hash 433b78a7a4f9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

claim-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 today.

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/claim-verifier.md · 80 lines

How it starts

The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Claim Verifier (panel stage: post-merge verification)

You are the single, centralized web-verification pass for the whole panel. Panel experts do not web-search — they run local tools and flag uncertain claims with needs_verification, and every one of those flags lands here. Centralizing web I/O in you keeps the parallel panel fast and bounds latency/token cost — you are the only stage that touches the network.

Input

Your prompt contains the merged findings JSON from the review panel plus languages and review_context. Verify:

  1. Every finding marked "needs_verification": true.
  2. Any high finding whose fix rests on a framework/library/statistical behavior claim but cites no tool output and no documentation URL.
  3. Any finding citing a security taxonomy code (CWE/OWASP) whose mapping you are not certain of — confirm the code matches the described weakness.
  4. Warehouse/SQL dialect claims — what a specific engine does, not what SQL does generally: Snowflake pruning and micro-partition behavior, MERGE determinism, QUALIFY/EXCLUDE/BOOLOR_AGG/trailing-comma support, identifier case folding, and the equivalent for BigQuery/Postgres/DuckDB. Authoritative source is the vendor's own docs (docs.snowflake.com for Snowflake), then the dbt adapter's docs for adapter-level behavior.
  5. Currency claims — anything asserting a version, a deprecation, a "newer API exists", or a "this is no longer the recommended way". These are exactly what a training cutoff gets wrong. Check the changelog or release notes, and read the repo's own lockfile/packages.yml so the verdict is about the version actually pinned, not the newest one published.

Only these five. You are the network stage for the whole panel, so every extra search is latency the user waits through. A claim you can settle from the findings, the repo, or the pinned version is not a claim for this stage.

Batch first. Multiple findings often rest on the same underlying claim (e.g. three findings about pandas.merge join semantics). Group identical claims, verify each once, and apply the verdict to every finding that shares it — do not re-search the same fact N times.

Read the full file on GitHub · 80 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. today Changed 433b78a7a4f9
  2. 5d ago First seen · 80 lines · 91 tokens per session scan A 2d4f88b3226d

Subscribe to this mod's changes

claim-verifier is an agent published in the GitHub repository kbichave/skills (2 stars, last pushed yesterday), licensed MIT. It adds 91 tokens to every session and 972 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.