Verification Engineer — Development agent for Claude Code
Independent verification engineer that reproduces reported behavior, executes acceptance, regression, and failure-path checks in a disposable worktree, and returns a pass/fail/inconclusive verdict wit.
How to install Verification Engineer
Installs to ~/.claude/agents/latent-sre-sde-agents-verification-engineer.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/latent-sre/sde-agents/HEAD/agents/verification-engineer.md -o ~/.claude/agents/latent-sre-sde-agents-verification-engineer.md Restart Claude Code, or start a new session, for it to be picked up.
What Verification Engineer does
name: verification-engineer description: Independent verification engineer that reproduces reported behavior, executes acceptance, regression, and failure-path checks in a disposable worktree, and returns a pass/fail/inconclusive verdict with evidence bound to the exact revision tested — authoring missing tests, never touching product code. Use for "verify the fix actually works", "independently confirm this bug is gone", "run the acceptance checks", or reproducing reported behavior against
Alternatives in Development
- Code Refactorer — Use this agent when you need to improve existing code structure, readability, or maintainability wit 2k ★
- Correlator — Finding correlation engine 812 ★
- Checklist Verifier — Verifies a single claim from the verification checklist against the actual source code 115 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Prompt Executor
Executes a prompt in an isolated worktree environment and captures results. Use when worktree path and prompt
Flutter Debugger
Use when investigating bugs, crashes, unexpected behavior, or flaky tests. Reproduces, isolates, and proposes
Disposable Reader
One-shot reader. Opens long files, logs, suite output, or documentation and returns a summary of at most 30 li
Hcs Eval Reviewer
Reviews regression-trap quality and eval harness coverage for HCS. Ensures traps capture real failure classes
Squad Fix
Operate as the squad's Bugfix Controller — reproduce a concrete failure, prove its root cause and blast radius
Background Reviewer
Use this agent as the asynchronous deep reviewer that runs off the hot path. Use for routine code review, depe
Related Skills
LinkedIn Post Analyser
Claude Code / Cowork skill that scores a drafted LinkedIn post across 7 platform dimensions before you publish
Chess Verdict
Claude skill that gives a verdict on a chess position: reads a board diagram or FEN, runs Stockfish, and retur
Cmd Judge
Adversarial verifier for the PROVE phase: treats a finished-work report as untrusted claims, re-runs every ver