Tribunal — Usage Guide

Tribunal is a Claude Code quality enforcement plugin. Once installed (pip install tribunal && tribunal init), it hooks into every file write, test run, and agent interaction.

Core Commands

tribunal status

Show current config, hook health, and audit summary.

tribunal status
tribunal status --json

tribunal doctor

Pre-flight health check — all hooks, deps, .claude/ wiring.

tribunal doctor        # interactive
tribunal doctor --json # for CI

Exit code 1 = something broken. Follow the ✗ suggestions to fix.

tribunal ci — Standalone quality gate

Run all checks without Claude Code. Use in CI pipelines.

tribunal ci                           # check git-diff'd files
tribunal ci src/payments.py           # specific files
tribunal ci --json                    # machine-readable report
tribunal ci --coverage-threshold 80   # fail if coverage < 80%

Exit codes: 0 = all pass · 1 = failures · 2 = warnings only

tribunal init — Project setup

tribunal init                        # interactive wizard
tribunal init --non-interactive      # CI/agent use, sensible defaults
tribunal init --project-dir /path    # target specific directory

Quality Modes

Control enforcement level per session via CLAUDE_CODE_MODE env var:

Add a "tribunal" block to any mode JSON:

{
  "tribunal": {
    "tddEnforcement": "advisory",
    "secretScanning": true,
    "coverageThreshold": 70,
    "blockOnFail": false
  }
}

Hook Events

Tribunal intercepts these Claude Code lifecycle events:

Plugin Packs

tribunal list-packs                    # browse registry
tribunal install python-strict         # 90% coverage, type hints, docstrings
tribunal install go-tdd                # go test -cover, go vet, -race flag
tribunal install nextjs-quality        # TypeScript strict, component tests, a11y
tribunal install https://github.com/org/custom-pack   # direct URL
tribunal install ./local-pack/         # local directory

MCP Integration

Tribunal exposes MCP tools queryable by any Claude Code session:

tribunal_status       — current hook config, mode, version
tribunal_audit        — recent audit entries (filterable by outcome/agent)
tribunal_check_file   — run file_checker on any path
tribunal_agent_summary — quality summary for a specific agent_id

Use via Claude Code's MCP connector pointing to the Tribunal MCP server.

Agent Teams Quality Gates

In multi-agent workflows, Tribunal gates quality between agents:

Orchestrator → spawns Sub-agent A
Sub-agent A writes files
TeammateIdle fires → tribunal checks A's files
  ✅ pass → orchestrator continues
  ❌ fail → orchestrator blocked until A fixes issues

Configure in tribunal/rules/multi-agent-quality.md (auto-injected).

Audit Log

All hook events logged to .tribunal/audit.jsonl:

cat .tribunal/audit.jsonl | jq '.[] | select(.outcome=="blocked")'

Fields: timestamp, hook_name, file_path, outcome, duration_ms, agent_id, detail

Context Monitor Thresholds

Configure in tribunal/settings.json:

{
  "contextMonitor": {
    "warnThreshold": 80,
    "handoffThreshold": 90
  }
}

Set CLAUDE_CONTEXT_WINDOW=1000000 env var for 1M context window (thresholds auto-scale).

VS Code Extension

After install, open VS Code in your project. The Tribunal extension:

• Shows 🏛 Passing / 🏛 1 Warning / 🏛 2 Blocked in status bar
• Adds gutter icons (✅/⚠️/⛔) per file after hooks fire
• Streams hook events to the "Tribunal" output channel
• Reads .tribunal/audit.jsonl automatically

Live Dashboard

# Open in browser after starting a Claude Code session
open tribunal/ui/viewer.html

Shows: live hook feed, context gauge, session stats, filterable audit log.

Full docs: https://tribunal.dev · GitHub: https://github.com/thebotclub/tribunal