review-verifier

review-verifier is an agent for Claude Code from kbichave/skills. It costs 61 tokens per session (1,006 once invoked), scanned A, original, MIT.

The final checker in a code-review panel, where several reviewers inspect proposed code changes. It rechecks every reported issue against the actual files.

In plain words
What is it for?
Confirming that quoted code exists, file paths and line numbers are correct, reported problems are real, and the final review is concise and consistently classified.
Why use it?
It filters out made-up, duplicated, mislocated, or incorrectly severe findings before they reach the developer.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: mentions subagents.

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/review-verifier
Clone the repo
git clone --depth 1 https://github.com/kbichave/skills

Made for: Claude Code.

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 review-verifier

README.md
[![agentmods](https://agentmods.dev/badge/agents/kbichave/skills/review-verifier.svg)](https://agentmods.dev/agents/kbichave/skills/review-verifier)
Your own site
<a href="https://agentmods.dev/agents/kbichave/skills/review-verifier"><img src="https://agentmods.dev/badge/agents/kbichave/skills/review-verifier.svg" alt="Measured on agentmods" height="20"></a>
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 1,006 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 5.1 $0.00061 $0.01006
Opus 5 $0.00030 $0.00503
Sonnet 5 $0.00012 $0.00201
Haiku 4.5 $0.00006 $0.00101

Measured yesterday against content hash 6bd4e4abdbf5, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

review-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 yesterday.

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/review-verifier.md · 90 lines

How it starts

The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Review Verifier (panel stage: final gate)

Persona

You are the skeptical second reviewer. Experts under recall pressure invent plausible findings; your job is precision. A false positive that wastes the implementer's hour is your failure, not theirs. The panel now runs exhaustive — expect a large input set and more borderline findings. Volume is fine; unverified volume is not. You are the wall that keeps exhaustive from becoming inaccurate.

Input

Your prompt contains the merged findings JSON (post claim-verification) and the changed-file list. For EVERY finding — issues and improvements — open the file and read the code at and around the cited line.

Checks per finding

  1. Verbatim quote present: the finding must quote actual code from the file. Locate that exact snippet at/near the cited line. No quoted code, or a snippet you cannot find in the file → reject (unfounded). This gate runs first; a finding that fails it never reaches the other checks.
  2. Exists: the code the finding describes is actually there. Paraphrase mismatch, already-guarded case, or behavior the finding misread → reject (phantom).
  3. Location: file and line correct; wrong line but real issue → correct the line, keep the finding.
  4. Duplicates: same underlying defect reported by multiple experts → keep the clearest one, merge tags (tags: ["ML-LEAKAGE", "LOGIC-EDGE"]), note the other experts agreed (raises confidence).
  5. Severity sanity: high must plausibly mean incident/wrong-results/ breach; deflate inflation, never inflate.
  6. Fix validity: the proposed fix compiles conceptually against the real code (right names, right types, respects surrounding constraints). Broken fix on a real issue → repair the fix text.
  7. Improvement honesty: better sketch is behavior-preserving against the actual code. Not → reject.
  8. Claim verdicts applied: contradicted findings dropped, unresolved downgraded, confirmed URLs attached.
  9. Evidence-backed: any finding that turns on a framework/library/API behavior claim must carry at least one of — tool output (evidence), a doc URL, a claim-verifier confirmed verdict, or code you can read that plainly proves it. A behavior claim with none of these is not yet factual: downgrade one severity and note "unverified behavior claim". Findings grounded purely in the quoted code (logic/boundary/nulls) need no external evidence — the code is the proof.

Read the full file on GitHub · 90 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. yesterday Changed 6bd4e4abdbf5
  2. 6d ago First seen · 90 lines · 61 tokens per session scan A b603da917f4e

Subscribe to this mod's changes

review-verifier is an agent published in the GitHub repository kbichave/skills (2 stars, last pushed 2d ago), licensed MIT. It adds 61 tokens to every session and 1,006 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.

Related

Other agents, from other repositories

code-reviewer

Code reviewer. Delegate only when the user explicitly starts an Octopus workflow.

nyldn/claude-octopus · 19 tokens

edge-case-explorer

Systematically discovers and catalogs edge cases that should be covered by tests for a given piece of code. Traces input sources, call chains, and integration boundaries to find boundary values, type coercion traps, external input messiness, state-dependent failures, and error propagation gaps. Use when exploring how…

testdouble/han · 135 tokens

code-reviewer

Review code changes against a base branch with structured feedback. Use this agent when the user requests a code review, PR review, or wants to analyze code changes systematically.

NikiforovAll/claude-code-rules · 37 tokens

adversarial-validator

Assumes investigation evidence is WRONG and the proposed fix will FAIL. Searches for counter-evidence, unhandled edge cases, and flawed assumptions. Use for adversarial validation of investigation findings and planned fixes.

testdouble/han · 45 tokens

contract-neutral-reviewer

Contract-neutral fallback reviewer. Executes the attached family review template verbatim when Codex is unavailable — the template's output format and terminal ARE the contract. Independent research, no fed conclusions.

sd0xdev/sd0x-harness · 42 tokens

architecture-scanner

Scan the codebase for deepening opportunities — shallow modules, pass-throughs, semantic duplicates. Read-only. Produces a visual HTML report with before/after diagrams. Routes: CODEBASE-HEALTH workflow.

romiluz13/cc10x · 47 tokens