Coherence Judge — Development agent for Claude Code
Truth Tribunal judge — internal consistency (sums, dates, entities).
How to install Coherence Judge
Installs to ~/.claude/agents/gonzalezpazmonica-pm-workspace-coherence-judge.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/gonzalezpazmonica/pm-workspace/HEAD/.claude/agents/coherence-judge.md -o ~/.claude/agents/gonzalezpazmonica-pm-workspace-coherence-judge.md Restart Claude Code, or start a new session, for it to be picked up.
What Coherence Judge does
name: coherence-judge description: Truth Tribunal judge — internal consistency (sums, dates, entities) model: opencode-go/deepseek-v4-flash permission_level: L1 tools: [Read, Bash] token_budget: per_invocation: 60000 context_window_target: 8500 escalation_policy: escalate max_context_tokens: 8000 output_max_tokens: 600
Coherence Judge — Truth Tribunal
You are one of 7 judges in Savia's Truth Tribunal (SPEC-106). Your focus: **internal consistency**. A report can be fully source
Alternatives in Development
- Activity Log Expert — Use this agent proactively when working with PostHog's comprehensive activity logging system, including implem 39.2k ★
- Tend — Tend the Allium garden 3.7k ★
- Review Agent — You are a specialized review agent 3.6k ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Calibration Judge
Truth Tribunal judge — confidence statements match evidence strength
Completeness Judge
Truth Tribunal judge — report covers what its title/abstract promises
Cb Truth
CoalBoard's formal lens (sub2) — reasons from first principles, invariants, and proof; checks internal consist
Debate Tribunal
Multi-agent structured debate protocol where 3-5 persona-driven debaters argue positions through claim-evidenc
Dotnet Style Reviewer
Read-only review of Contoso Claims conventions — controller/service/DbContext layering, DTOs vs entities at th
Verify Architecture
STAGE 4 VERIFICATION - Architectural coherence. Ensures code follows established patterns, validates layering,
Related Skills
Ground Truth Evals
An LLM eval harness graded by computed ground truth, not an LLM judge. Worked example: poker, served to models
Tribunal
Convene the blind tribunal on the current diff — cross-family jurors, one lens each, author-redacted envelope,
Codebase Tribunal
⚖️ Your codebase is on trial: 5-8 AI expert personas (PM/architect/backend/security/QA/DevOps) audit any proje