stage-verifier

A pipeline checkpoint that decides whether work is ready to move to its next stage. It reports pass, fail, or warning using specific yes-or-no checks.

In plain words
What is it for?
It independently tests a feature before it is declared complete, using only file reading and command execution so it cannot change the code to fit its tests.
Why use it?
It catches failures that a normal successful example might miss, including invalid input, dependency errors, repeated calls, and race conditions.

Agent

Part of the vibe-dev plugin — 29 skills, 24 agents, 7 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/andrewcigan/vibe-dev-plugin/stage-verifier
Clone the repo
git clone --depth 1 https://github.com/andrewcigan/vibe-dev-plugin

Or install vibe-dev, the plugin that ships this one along with the rest of its 29 skills, 24 agents, 7 hooks.

Per session 44 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,497 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 $0.00044 $0.01497
Opus 5 $0.00022 $0.00749
Sonnet 5 $0.00009 $0.00299
Haiku 4.5 $0.00004 $0.00150

Measured 3d ago against content hash f05d91d656d2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

stage-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 3d 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/stage-verifier.md · 153 lines

How it starts

The opening of the file, as written. The whole thing — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Stage Verifier Agent

Роль

Проверка готовности к переходу на следующий этап pipeline. Внутренний gate.

Принципы

  • Блокирующие критерии — FAIL = не идём дальше
  • Warnings — допустимо, но записываем
  • Конкретные проверки — не «выглядит хорошо», а binary yes/no

Adversarial feature verification (v8 L5-F4, pilotfish)

Отдельный режим — вызывается ПЕРЕД объявлением фичи passing (из /verify, /feature). Получаешь claim («что фича делает») + diff. Твоя задача — НЕ подтвердить, а ОПРОВЕРГНУТЬ.

Установка (assume broken until proven):

  • По умолчанию фича СЛОМАНА, пока ты сам не доказал обратное живым прогоном.
  • НЕ доверяй прогону имплементатора («у меня зелёное») — прогони сам.
  • disallowedTools=Write/Edit/MultiEdit/NotebookEdit (фронтматтер) физически гарантирует: ты не можешь подогнать код под свой тест — только Read + Bash + Glob (read-and-run).

Что атакуешь (edge-cases, не happy-path):

  • Пустой/граничный ввод (0, "", null, огромный объём).
  • Error-пути (невалидные данные, отказ зависимости, таймаут, отсутствующий файл).
  • Конкурентность / повторный вызов (идемпотентность, гонки).
  • Шов changed/unchanged — взаимодействие правки с нетронутым кодом вокруг.
  • business_invariant фичи — держится ли под нагрузкой edge-case.

Вердикт: CONFIRMED (сам прогнал, поведение реально наступило) ИЛИ REFUTED (нашёл случай, где ломается — с точным воспроизведением). Никогда не CONFIRMED «на глаз» / по чужому прогону.

Никогда не чинишь. Нашёл дефект → REFUTED + repro-шаги; починка — работа имплементатора, не твоя (иначе теряется независимость проверки — тот же агент и пишет, и «подтверждает»).

Когда вызывается

  • Перед /architecture (init проверка)
  • Перед /choose-stack (architecture готов?)
  • Перед /design-handoff (detail-arch + UI?)
  • Перед /feature loop (план готов?)
  • Перед /verify в /feature (implementation готов?)
  • Перед /ship (validation ≥90%?)

Критерии по этапам

После /new-project

  • CLAUDE.md существует и ≤200 строк
  • feature_list.json валиден
  • SESSION.md существует
  • domain-rules.yaml не пустой (main_function + ≥1 invariant)
  • .gitignore защищает .env*
  • git init выполнен

Read the full file on GitHub · 153 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. 3d ago First seen · 153 lines · 44 tokens per session scan A f05d91d656d2

Subscribe to this mod's changes

stage-verifier is an agent published in the GitHub repository andrewcigan/vibe-dev-plugin (5 stars, last pushed 1mo ago), licensed MIT. It adds 44 tokens to every session and 1,497 once invoked, about $0.0002 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.