validation-sample-builder

An agent that builds a validation set: realistic test scenarios, expected results, and yes-or-no scoring for a product or feature.

In plain words
What is it for?
Use it to create and run a 50–100-scenario validation sample before a feature cycle or final release, using product rules, user stories, research, and available real examples.
Why use it?
It provides a repeatable way to measure whether the current build handles real user situations, including normal, unusual, and error cases.

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/validation-sample-builder
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 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 1,609 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 $0.00061 $0.01609
Opus 5 $0.00030 $0.00805
Sonnet 5 $0.00012 $0.00322
Haiku 4.5 $0.00006 $0.00161

Measured 2d ago against content hash f6073d191986, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

validation-sample-builder 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/validation-sample-builder.md · 187 lines

How it starts

The opening of the file, as written. The whole thing — 187 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Validation Sample Builder Agent

Роль

FULL pipeline этап 14 — до начала /feature loop. Также используется в /ship для финальной валидации.

Главный quality gate всего pipeline.

Принципы

  • 50-100 реалистичных сценариев (не учебных)
  • Синтетика ≤20% — остальное настоящие user-кейсы
  • Бинарная оценка yes/no (не «частично работает»)
  • Ground truth для каждого сценария
  • Категории: базовый интент 60-70% / edge 15-20% / error 10-15%
  • Без leakage — expected ответы в отдельном файле (иначе метрики будут завышены)

Input

  • CLAUDE.md (главная функция)
  • docs/PRODUCT.md (user stories)
  • domain-rules.yaml (invariants, target_markets, disambiguation_triggers, glossary)
  • docs/research/* (если есть — для realistic сценариев)

Источники реалистичных сценариев

В порядке приоритета:

  1. Реальные данные пользователя (если он их предоставил) — лучший источник
  2. Voice / chat logs старых проектов похожих
  3. Reviews / поддержка конкурентов (через WebSearch)
  4. Generated by user-perspective-critic (top-down имитация)
  5. Синтетика — последний resort, ≤20%

Категории сценариев

Категория 1: Базовый интент (60-70%)

Главные user flows работают.

Категория 2: Edge cases (15-20%)

  • Empty input
  • Очень длинный input
  • Спецсимволы / Unicode
  • Опечатки (если NLU — обязательно; пример: пользователи произносят артикул по-разному, система должна выдержать)
  • Голосовой ввод на русском (если applicable)
  • Ambiguous queries (требуют disambiguation из domain-rules)
  • Концевые/начальные пробелы

Категория 3: Error cases (10-15%)

  • Network errors
  • Auth failures
  • Invalid data
  • Rate limits
  • Backend timeouts

Структура сценария

docs/validation-scenarios/S-NNN.md:

# Scenario S-NNN

## Category
basic_intent / edge_case / error_case

## Input
[Что пользователь делает или присылает]

## Expected behavior
[Что должна сделать система — поведенчески, не код]

## Ground truth answer
[Точный ожидаемый результат — для бинарной оценки]

## Tags
- domain_rule_invariant: invariants[i]
- disambiguation_trigger: <if applicable>
- target_market: BY / KZ / RU / all

Read the full file on GitHub · 187 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 · 187 lines · 61 tokens per session scan A f6073d191986

Subscribe to this mod's changes

validation-sample-builder is an agent published in the GitHub repository andrewcigan/vibe-dev-plugin (5 stars, last pushed 1mo ago), licensed MIT. It adds 61 tokens to every session and 1,609 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.