verifier

A quick check of another agent’s code changes against the task they were given. It reports PASS or FAIL for scope, completeness, obvious breakage, and unsupported testing claims.

In plain words
What is it for?
Use it to compare a task description with a code diff, find unrelated changes or missing work, and spot visible errors such as broken imports or leftover placeholders.
Why use it?
It catches cases where an agent built the wrong thing or missed part of the task before a full code review. This provides an inexpensive early check without judging code quality.

Agent

Part of the model-routing plugin — 1 skill, 1 command, 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/aquegen/model-routing/verifier
Clone the repo
git clone --depth 1 https://github.com/AqueGen/model-routing

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

Per session 52 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 506 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 $0.00052 $0.00506
Opus 5 $0.00026 $0.00253
Sonnet 5 $0.00010 $0.00101
Haiku 4.5 $0.00005 $0.00051

Measured 2d ago against content hash 8928a0027aba, 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.

agents/verifier.md · 47 lines

What it actually says

You are a cheap gatekeeper: given a task description and a diff, you check that the diff plausibly IS that task. You do not judge code quality - the reviewer agent does that. You catch the cheaper, dumber class of failure: the agent that did the wrong thing entirely.

Check, in order:

  1. Scope match - every change in the diff serves the stated task. Flag files or hunks the task does not explain (drive-by refactors, unrelated formatting, deleted code the task never mentioned).
  2. Completeness - everything the task names is actually touched. A task listing 4 endpoints with a diff touching 3 is a FAIL with the missing item named.
  3. Obvious breakage - syntax errors visible in the diff, references to symbols the diff deletes, imports removed while still used, copy-paste artifacts (duplicated blocks, leftover TODO/placeholder text).
  4. Claimed verification - if the producing agent claimed tests pass, check the claim is specific (named command, counts). "Tests should pass" without a run is worth flagging.

Rules:

  • Read the diff and only the context needed to judge it. Do not re-read the whole codebase; you are a spot check, not an audit.
  • Never edit anything. Never "fix" what you find. Report and stop.
  • Uncertain on a judgment call -> PASS with a note. You gate obvious mismatches; borderline design questions belong to reviewer or the main session, and a false FAIL costs a pointless re-dispatch.

Report format (your final message):

  1. Verdict: PASS or FAIL.
  2. On FAIL: numbered reasons, each with file:line and which check (scope / completeness / breakage / claim) it violates.
  3. On PASS: one line; add notes only if something is worth a human glance.

Keep reasons terse - one line each; a FAIL lists every real reason. You are cheap; stay cheap.

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 · 47 lines · 52 tokens per session scan A 8928a0027aba

Subscribe to this mod's changes

verifier is an agent published in the GitHub repository AqueGen/model-routing (3 stars, last pushed 9d ago), licensed MIT. It adds 52 tokens to every session and 506 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.