Sonmat Witness — Development agent for Claude Code
External witness agent.
How to install Sonmat Witness
Installs to ~/.claude/agents/jun0-ds-sonmat-sonmat-witness.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/jun0-ds/sonmat/HEAD/agents/sonmat-witness.md -o ~/.claude/agents/jun0-ds-sonmat-sonmat-witness.md Restart Claude Code, or start a new session, for it to be picked up.
What Sonmat Witness does
name: sonmat-witness description: External witness agent. Verifies intent-artifact match using user turn cascade and ground truth. Protocol-isolated from main reasoning — see §Isolation stack for what "isolated" actually means on current Claude Code. tools:
- Read
- Bash
- Grep
- Glob
Role
External witness. Your only job is to find discrepancies between **what the user asked for** and **what the artifact actually is**. You do not reason. You compare.
You are spawned by auto
Alternatives in Development
- Eval Engineer — GAIA evaluation framework specialist 1.5k ★
- Chain Executor — Executes delegated chain steps with context isolation 186 ★
- Retro Reviewer — Reviews one prior session through ax's normalized cross-harness Turn view and emits a structured retro (tried 103 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Cutter
Turn recorded biographies into the shaped scene-list of a novel — the documentary editor's cut. Given the exha
Contract Diff Auditor
Adversarial review of a contract diff -- identifies clauses that shift risk, limit rights, or create obligatio
Sme Eval Triage
Triage a golden-set failure from the compliance-SME seat before anyone edits ground truth. Use whenever make e
Stroi Explorer
Use this agent when a planner or skill needs ground truth about the codebase before designing against it. Typi
Doc Analyser
Per-page documentation analyser for the ground-truth doc-lineage layer. Reads one published doc page (the GitB
Issue Planner
Use this agent to turn a triaged issue into an approved implementation plan with formal acceptance criteria. T
Related Skills
Rails AI Context
45 MCP tools that give AI coding agents ground truth about your Rails app: schema, models, routes, controllers
Claude Boot Stats
Per-component token cost analyzer for Claude Code first-turn context: HTTP intercept proxy + Qwen3 BPE + groun
Closeout
Post-project audit. Diffs SCOPE.md deliverables against ground truth using tier-aware verification (code-done