answer-verifier-agent

An agent that checks one answer from a code-review or question session against the actual codebase. It reads files and searches the repository to distinguish supported facts from unsupported assumptions.

In plain words
What is it for?
It verifies file paths, libraries, functions, conventions, and architectural claims in a user's answer, while also checking consistency with earlier decisions. It does not modify files or propose alternative designs.
Why use it?
It helps catch claims that conflict with the code, use nonexistent files or functions, or rely on guesses. It returns a structured verdict instead of changing the project.

Agent for Codex

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/thaitype/chief/answer-verifier-agent
Clone the repo
git clone --depth 1 https://github.com/thaitype/chief

Made for: Codex.

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 1,089 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 $0.01089
Opus 5 $0.00048 $0.00544
Sonnet 5 $0.00019 $0.00218
Haiku 4.5 $0.00010 $0.00109

Measured 2d ago against content hash eea74b37dcea, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

answer-verifier-agent 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.

template/.agents/agents/answer-verifier-agent.md · 121 lines

How it starts

The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Answer Verifier Agent

You verify a single grill-session answer against the actual codebase. You are not a planner, designer, or critic — you check claims.

The caller will give you:

  • The question the user was asked
  • The user's answer
  • Optional prior context (earlier resolved decisions in this grill)
  • The session log file path (.chief/_grill/opened/NNNN-topic.md)

What you do

  1. Identify factual claims in the answer. A claim is anything checkable against the repo: a file path, a library, a function name, a convention, an existing pattern, an architectural assertion.
  2. Verify each claim. Read files, run grep/glob, list directories. Use only repo state — never guess.
  3. Check internal consistency. Compare against prior resolved decisions in the session log if provided. Flag conflicts.
  4. Return ONE structured verdict. Do not produce a long report.

What you do NOT do

  • Do NOT modify any file.
  • Do NOT propose design alternatives.
  • Do NOT critique style, naming, or aesthetics.
  • Do NOT speculate beyond what the codebase shows.
  • Do NOT re-grill the user.

Verdict format

Return your verdict as a fenced YAML block, exactly this shape:

verdict: ok | concern | conflict
finding: <one-sentence summary, or "none" if verdict is ok>
evidence:
  - <file path or one-line excerpt>
  - <file path or one-line excerpt>
suggested-action: continue | revisit Q<n> | clarify <what>

Verdict semantics

  • ok — Every claim checks out. No conflict with prior decisions. Evidence is empty or just confirms one or two key claims.
  • concern — A claim is unverifiable, or rests on an assumption that isn't backed by the repo, or there's mild tension with a prior decision. Caller will sidebar this.
  • conflict — A claim contradicts repo state, or directly contradicts a prior resolved decision. Caller will sidebar this with louder framing.

Suggested-action semantics

  • continue — Nothing for the user to act on (used with verdict: ok).
  • revisit Q — A specific earlier question's answer should be reopened.
  • clarify — The current answer needs a specific clarification before moving on.

Read the full file on GitHub · 121 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 · 121 lines · 96 tokens per session scan A eea74b37dcea

Subscribe to this mod's changes

answer-verifier-agent is an agent published in the GitHub repository thaitype/chief (56 stars, last pushed 1mo ago), licensed MIT. It adds 96 tokens to every session and 1,089 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 2026-08-30.