Tribunal

Heading to v3 -- the open audit and policy layer for coding agents

Tribunal is being rebuilt around a single problem: every team running Claude Code, Cursor, Copilot CLI, and Codex CLI has four separate audit logs, four separate spend caps, and zero cross-agent policy. The v3 line ships a local daemon + normalised event schema + policy engine that sits underneath all of them. OSS-core (MIT), self-hostable, with a paid hosted tier and a compliance tier.

The v2 quality-gate features documented below still work -- they're moving to a tribunal scan subcommand in 3.0 and remain supported. If you're here for CI checkers, you're in the right place.

Target: 3.0.0a1 in week 2, GA in week 9 of the v3 plan.

Quality gates for AI-generated code. v2.0.1

Tribunal scans code for secrets, enforces TDD, lints Python / TypeScript / Go, and outputs SARIF -- in one command. Works everywhere: CI pipelines, pre-commit hooks, local dev. Agent-agnostic.

5 checkers * SARIF 2.1.0 * GitHub Action * pre-commit hook

Quick Start

pip install tribunal
tribunal ci .

That's it. Tribunal walks every source file, runs applicable checkers, and exits non-zero if anything fails.

What It Does

Secrets scanning runs on every file. Language checkers run only on matching extensions.

Output Formats

tribunal ci .                    # Human-readable text (default)
tribunal ci . --format sarif     # SARIF 2.1.0 for GitHub Code Scanning
tribunal ci . --format json      # Machine-readable JSON
tribunal ci . --output report.sarif  # Write to file

CLI Commands

tribunal ci .               # Run all checkers on current directory
tribunal ci src/ tests/      # Check specific paths
tribunal ci . --checkers secrets,python  # Run only specific checkers
tribunal ci . --format sarif --output results.sarif  # SARIF output

tribunal init                # Set up project config
tribunal status              # Show active rules and config
tribunal rules               # List configured rules
tribunal audit               # View audit log
tribunal config              # Show resolved config
tribunal pack list           # Show available rule packs
tribunal doctor              # Health check

GitHub Action

# .github/workflows/tribunal.yml
name: Tribunal CI
on: [push, pull_request]

jobs:
  tribunal:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: thebotclub/tribunal.dev/tribunal@v2.0.1

The action installs Tribunal, runs tribunal ci . with SARIF output, and uploads results to GitHub Code Scanning automatically.

pre-commit Hook

# .pre-commit-config.yaml
repos:
  - repo: https://github.com/thebotclub/tribunal.dev
    rev: v2.0.1
    hooks:
      - id: tribunal-ci        # Full check suite
      - id: tribunal-secrets   # Secrets only (fast)

Secrets Scanning

14 built-in patterns covering:

• AWS access keys and secret keys
• GitHub personal access tokens (classic and fine-grained)
• Anthropic and OpenAI API keys
• Slack tokens and webhooks
• Private keys (RSA, EC, etc.)
• Database connection URLs with passwords
• JWTs, Bearer tokens
• Generic hex secrets and API keys

.secretsignore

Suppress false positives with a .secretsignore file in your project root:

# Patterns (one per line, matched against file path)
docs/examples/*
test_fixtures/mock_keys.py

Placeholder Detection

Tribunal automatically skips placeholder values like your-api-key-here, CHANGE_ME, xxxx, and TODO patterns -- only real secrets trigger findings.

TDD Enforcement

For every source file, Tribunal checks whether a corresponding test file exists:

Files that are reasonably excluded (test files themselves, __init__.py, index.ts, main.go) are skipped.

Rule Packs

Pre-built rule sets for common standards:

tribunal pack list           # Show available packs
tribunal pack install soc2   # Install SOC 2 rules

Available: soc2, startup, enterprise, security.

Configuration

Tribunal reads config from .tribunal/config.yaml:

rules:
  tdd-python:
    match:
      path: "*.py"
    action: block
    condition: no-matching-test
    message: "Write a failing test first."

  no-secrets:
    action: block
    condition: contains-secret
    message: "Possible secret detected."

Programmatic API

from tribunal.checkers import run_checkers, collect_files
from tribunal.sarif import findings_to_sarif, sarif_to_json

files = collect_files("/path/to/project")
results = run_checkers(files, project_root="/path/to/project")

# Check pass/fail
passed = all(r.passed for r in results)

# Generate SARIF
sarif = findings_to_sarif(results, "/path/to/project")
print(sarif_to_json(sarif))

REST endpoints: /api/projects, /api/summary, /api/projects/{id}/audit|cost|agents.

VS Code Extension

Visual governance in the editor sidebar:

• Rules Tree -- See all rules with action icons
• Audit Tree -- Browse recent events
• Cost Tree -- Track budget usage
• Agents Tree -- Monitor sub-agents
• Status Bar -- Rule count and block count at a glance

Architecture

.tribunal/
??? rules.yaml          # Rule definitions
??? config.yaml         # Project configuration
??? permissions.yaml    # Permission policies
??? audit.jsonl         # Audit log (gitignored)
??? state.json          # Cost tracking state (gitignored)
??? skills/             # Custom skills
??? bundle.json         # Air-gapped bundle (export)

.claude/
??? claudeconfig.json   # Hook wiring for Claude Code
??? memory/             # Tribunal memory entries
    ??? tribunal-rule-*.md
    ??? tribunal-session-*.md

License

MIT