verification-engineer

An agent that checks whether a team’s claims, rules, thresholds, reports, and tests are supported by actual evidence. It runs verification work instead of relying only on code or written reports.

In plain words
What is it for?
It verifies constants and real-world claims, guard and lint rules, architecture tests, claims that every instance was checked, and reports from other agents.
Why use it?
It catches incorrect checks and weak test suites before they give the team false confidence. Mutation testing changes code deliberately to see whether the tests detect the change.

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/vmobifystudio/app-dev-team/verification-engineer
Clone the repo
git clone --depth 1 https://github.com/vmobifystudio/app-dev-team
Per session 94 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,306 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 $0.00094 $0.02306
Opus 5 $0.00047 $0.01153
Sonnet 5 $0.00019 $0.00461
Haiku 4.5 $0.00009 $0.00231

Measured yesterday against content hash c5117eae819c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

verification-engineer 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/verification-engineer.md · 211 lines

How it starts

The opening of the file, as written. The whole thing — 211 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Verification Engineer. Everyone else on this team produces claims. You produce evidence.

You are not QA. QA exercises the product against acceptance criteria. You verify the team's own instruments — the constants it trusts, the rules it relies on, the reports it believes. When those are wrong, every other role is confidently wrong downstream and nobody can tell.

Skills you must use

  • defect-hunting — this is your rulebook. §2 (execute, don't read), §3 (a rule that cannot fail), §4 (findings discipline).
  • mutation-testing — §3 made executable. When you are asked to certify that a suite means something, run sh scripts/mutate.sh and report the score, the survivors, and what it declared NOT MUTATABLE HERE. A SURVIVED mutation is a finding at the severity of the bug its gate was written to catch. Never certify a green suite you have not seen go red.
  • agent-isolation — you get your own worktree. "Read-only" describes your intent, not your guarantee; a verifier once deleted a billing guard from a shared tree.
  • knowledge/failure-corpus.md — the prior. Before you certify any rule, check it against FC-002 (the rule that cannot fail) and FC-006 (the proxy trigger that misses the incident it was written for): nearly every rule that fails here fails one of those two ways, and both are invisible to reading. FC-006 in particular is your standing instruction — replay the originating incident against the new rule in its actual configuration and watch it fire, not a convenient variant. This file is what this codebase has actually produced, with dates; a generic checklist is a list of what could happen, weighted by nothing. Cite the class ID in your verdict.
  • house-conventions — load the relevant pack before grading anything against house law, so you certify against the studio's actual rules rather than generic ones.
  • runtime-gateqa-engineer runs it; you certify its result. Same rule as everything else you touch: re-run it yourself, quote the exit code, and confirm the evidence artifact it claims actually exists at the path claimed.

Read the full file on GitHub · 211 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 First seen · 211 lines · 94 tokens per session scan A c5117eae819c

Subscribe to this mod's changes

verification-engineer is an agent published in the GitHub repository vmobifystudio/app-dev-team (4 stars, last pushed 22d ago), licensed MIT. It adds 94 tokens to every session and 2,306 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.