validation-sample

A skill for building a reference test set of 50–100 realistic scenarios for checking an agent or product. It separates user inputs from expected answers to prevent answer leakage and requires at least 90% before shipping.

In plain words
What is it for?
Collecting validation scenarios, generating expected results, running a baseline against the current build, and recording validation results.
Why use it?
It provides repeatable test cases for basic, unusual, and error situations, making it easier to measure whether a build behaves correctly.

Skill for Claude CodeCodex

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 skills/andrewcigan/vibe-dev-plugin/validation-sample
Any agent
npx skills add andrewcigan/vibe-dev-plugin --skill validation-sample
Clone the repo
git clone --depth 1 https://github.com/andrewcigan/vibe-dev-plugin

Made for: Claude Code, Codex.

Per session 85 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 559 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 $0.00085 $0.00559
Opus 5 $0.00043 $0.00280
Sonnet 5 $0.00017 $0.00112
Haiku 4.5 $0.00009 $0.00056

Measured 2d ago against content hash 6084473b1b99, 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 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.

skills/validation-sample/SKILL.md · 56 lines

What it actually says

/validation-sample

Построение эталонной выборки через validation-sample-builder agent.

Что происходит

  1. Subagent validation-sample-builder (Sonnet) читает PRODUCT, ARCHITECTURE, domain-rules
  2. Identify источники реалистичных сценариев:
    • User-provided data (приоритет)
    • Voice/chat logs old projects
    • Competitor reviews (WebSearch)
    • User-perspective-critic generated
    • Synthetic (≤20%)
  3. Строит 50-100 scenarios в категориях:
    • basic_intent 60-70%
    • edge 15-20%
    • error 10-15%
  4. Leak prevention:
    • docs/validation-scenarios/inputs/ — только input
    • docs/validation-scenarios/ground-truth/ — expected (только judge видит)
  5. Запускает baseline run на текущей сборке (если есть код)

Output

  • docs/validation-sample.md — сводка
  • docs/validation-scenarios/inputs/*.md — 50-100 сценариев
  • docs/validation-scenarios/ground-truth/*.md — expected
  • ./validation-runs/run-<ts>.jsonl — результаты

Critical gotchas

  • Leak prevention: expected НЕ в одном файле с input
  • Judge contains rule: YES если expected appears anywhere в Got, не exact match
  • Truncate Got >500 chars запрещён

Pass thresholds

  • ≥90% → ✓ /ship
  • 80-89% → 🟡 /ship с warnings, failed → backlog
  • <80% → ❌ stop, 5 Why на failed scenarios

Дальше

  • Если pre-/feature loop — запоминаем baseline, /feature feat-001
  • Если в /ship — финальный gate

Cost cap

$3. Может занять до часа compute на judge.

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 · 56 lines · 85 tokens per session scan A 6084473b1b99

Subscribe to this mod's changes

validation-sample is a skill published in the GitHub repository andrewcigan/vibe-dev-plugin (5 stars, last pushed 1mo ago), licensed MIT. It adds 85 tokens to every session and 559 once invoked, about $0.0004 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.

Related

Other skills, from other repositories

harness-engineering-guide

Audit, design, and implement AI agent harnesses for any codebase. A harness is the constraints, feedback loops, and verification systems surrounding AI coding agents — improving it is the highest-leverage way to improve AI code quality. Three modes: Audit (scorecard), Implement (set up components), Design (full…

OdradekAI/harness-engineering-guide · 182 tokens

google-drive-sheets

Find, read, export, edit, and manage the user's Google Drive, Docs, Sheets, and Slides through per-user OAuth.

yc-software/qm · 31 tokens

github-gitlab

Work with GitHub and GitLab repositories through resident gh/glab/git auth on the agent computer.

yc-software/qm · 25 tokens

harness-creator

Build, audit, and improve harnesses that make AI coding agents reliable: AGENTS.md/CLAUDE.md instruction files, feature/state tracking, verification gates, scope boundaries, session handoff, memory persistence, context budgets, tool-permission safety, and multi-agent coordination. Use this whenever a coding agent is…

walkinglabs/learn-harness-engineering · 142 tokens

miniapp

Build a tiny interactive HTML playground only when someone asks to see, play with, or step through a mechanism.

yc-software/qm · 25 tokens

commit-push-pr

Commit selected local changes, push the branch, and create or update a GitHub pull request with BitFun attribution. Use when the user asks to 提交 PR、提代码、commit and push、开 PR、create a pull request, or wants a Claude Code-like one-command PR publishing flow from BitFun.

GCWing/BitFun · 68 tokens