Harness Prover — Development agent for Claude Code
Running-app verifier for harness eval loops.
How to install Harness Prover
Installs to ~/.claude/agents/leadgrowgtm-loop-engineer-harness-prover.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/LeadGrowGTM/loop-engineer/HEAD/.claude/agents/harness-prover.md -o ~/.claude/agents/leadgrowgtm-loop-engineer-harness-prover.md Restart Claude Code, or start a new session, for it to be picked up.
What Harness Prover does
name: harness-prover description: Running-app verifier for harness eval loops. Drives the live feature (browser, API, or CLI) and returns a binary PROOF verdict before Checker scores artifacts. Only spawned for running-app goals — static artifact goals skip directly to Checker. Returns works/broken with command output or screenshot evidence. tools: Read, Bash model: claude-sonnet-5
You are the Harness Prover. You are at depth level 3 (goal=0, planner=1, maker=2, prover=3).
Your role: e
Alternatives in Development
- Agent Refactorer — Principal Clean Code Specialist specializing in code simplification, performance tuning, and refactoring loops 957 ★
- Browser Stealth Agent — Stealth browser automation agent for targets behind Cloudflare, Akamai, Google, DataDome, or PerimeterX bot de 812 ★
- Eval Implementer — Implementer role of the D365FO agent eval loop 136 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Harness Checker
Fresh-context artifact evaluator for goal loops. Reads only final artifacts — never Maker reasoning or PROGRES
Lean Prover Opus
Escalation-tier Lean 4 + Mathlib prover for the lean-proof-agents repo. Dispatched when a Sonnet lean-prover r
Path Prover
Produces live evidence that every layer a lane touched actually works — boots the local stack or targets stagi
Harness Shipper
Fresh shipping agent for goal loops. Runs the no-mistakes validation pipeline exactly once after Checker PASS,
Lean Prover
Writes and repairs Lean 4 + Mathlib proofs in the lean-proof-agents repo. Search-first, 2-attempt iteration ca
Lens Rumelt
Read-only strategy-kernel lens: diagnosis, guiding policy, coherent actions; flags fluff, goals-as-strategy an
Related Skills
LinkedIn Post Analyser
Claude Code / Cowork skill that scores a drafted LinkedIn post across 7 platform dimensions before you publish
Helpful Content Review Skill
A Claude skill that reviews and scores any draft or live page against Google's helpful, reliable, people-first
Conductor Loop
Run the conductor-builder → conductor-verifier maker/checker pattern on a task.