Fresh Verifier — Testing & QA agent for Claude Code
Fresh-context Opus 5 verifier.
How to install Fresh Verifier
Installs to ~/.claude/agents/y0f-fable-orchestration-5.1-fresh-verifier.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/y0f/fable-orchestration-5.1/HEAD/agents/fresh-verifier.md -o ~/.claude/agents/y0f-fable-orchestration-5.1-fresh-verifier.md Restart Claude Code, or start a new session, for it to be picked up.
What Fresh Verifier does
name: fresh-verifier description: Fresh-context Opus 5 verifier. Run after a builder reports done. Checks work against the spec, runs tests, reports evidence. Never the builder. Does not fix. model: opus effort: high tools: Read, Grep, Glob, Bash maxTurns: 40 color: red
You verify. You do not fix; Bash is for running tests and checks only, never for writing files.
Input: the spec (or plan) and the claim of completion. Output: PASS or FAIL per requirement with evidence from this session
Alternatives in Testing & QA
- Test Wiring Auditor — 変更差分に対してテスト網が追随しているかを fresh-context で監査する read-only auditor 3.1k ★
- Fable Verifier — Cold verifier for finished fable deliverables 837 ★
- Post Gate Agent — Review building phase implementation against plan requirements and test coverage 243 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Merge Gate Verifier
Read-only verification of a branch before it merges. Use when a worker reports a lane done and you need to che
Missile Builder
Missile worker. Implements one scoped feature or fix end to end in an isolated git worktree (acceptance test f
Haiku Tester
QA agent running Haiku. Use to test a finished change against an acceptance checklist — runs the project's aut
Evaluate
Evidence-based auditor in a fresh context. Verifies every claim in one named artifact (a spec cut, a board's c
Premise Checker
Adversarially test one explicit consequential premise in fresh read-only context; return supported, revise, or
Fresh Eyes Reviewer
First-time reader perspective on a paper. Reads cold with no prior context — not the spec, not the lit review,
Related Skills
Goalpost
Your agent says “All tests pass!” The hidden tests say 28/34. goalpost makes Claude Code's /goal prove it: fro
Did You Check
Checks whether the Claude Code best practices were actually followed, and blocks the turn when they were not:
AI Loop Skill
Your agent says "done" and it is not. A gated multi-agent circuit for Claude Code: whoever implements does not