eagle-governance

name	eagle-governance
description	Install hook-based engineering governance for agent-led projects. Use when the user says: 'govern this project', 'install governance hooks', 'agent governance', 'context bloat guard', 'hook-based engineering governance', 'stop agents ignoring instructions', or asks to make AGENTS.md / CLAUDE.md enforceable through hooks.

Eagle Governance

Install a project-local governance runtime that keeps instruction files small and lets hooks enforce the important rules.

What this does

1. Adds a small governance policy file: .eagle-governance.json
2. Installs hook scripts for Claude Code and Codex when those targets are selected
3. Adds small managed routing blocks to AGENTS.md and/or CLAUDE.md
4. Configures Grok Build through its Claude Code compatibility layer and Antigravity through .agents/hooks.json
5. Verifies hook scripts, JSON wiring, and instruction shims
6. Partners with Eagle Mem when available for durable handoff mirroring and feature-verification visibility
7. Ships an optional Eagle Eval governance scenario pack for regression checks

Commands

eagle-skills govern status --target all
eagle-skills govern apply --target all --mode warn
eagle-skills govern verify --target all

Use --dry-run before applying when you want to preview writes:

eagle-skills govern apply --target all --mode warn --dry-run

Default policy

• Default mode is warn
• Hard blocks remain active for destructive commands, secret-file writes, outside-root writes, and high-risk context handoff
• Context bloat uses Claude's real context percentage when available, then Eagle Mem turn budget, then optional transcript thresholds. High risk produces .eagle-governance/handoff.md and blocks further implementation
• Transcript size and changed-file count are warnings by default; they do not force handoff unless configured
• Eagle Mem bridge defaults to auto: Governance works without Eagle Mem, but mirrors handoffs and surfaces pending feature verification when Eagle Mem is installed
• Eagle Eval is optional and never runs inside hooks; use the governance pack when you want regression tests for hook behavior
• AGENTS.md and CLAUDE.md remain shims; do not duplicate the full policy there

Workflow

1. Run eagle-skills govern status --target all
2. Run eagle-skills govern apply --target all --mode warn --dry-run
3. Run eagle-skills govern apply --target all --mode warn
4. Run eagle-skills govern verify --target all
5. For Codex, open /hooks and trust the new project hooks when prompted
6. For Claude Code, restart or resume the project session so new hooks and the context statusline load
7. If Eagle Mem is installed, confirm govern status reports the bridge as available

Output

Return a short PM-readable report:

Eagle Governance:
  Policy: installed
  Claude Code: hooks installed and wired
  Codex: hooks installed and wired; trust review may be required
  Eagle Mem: bridge available; handoff mirror enabled
  Eagle Eval: governance pack available; CLI optional
  Grok Build: Claude-compatible hooks installed and wired
  Antigravity: hooks installed and wired

Principles

• Hooks are control; instruction files are routing context
• Keep injected context tiny
• Treat each provider's hook output contract separately
• Let Eagle Mem own durable recall, orchestration lanes, and release-boundary feature verification
• Let Eagle Eval test governance behavior outside the live hook path
• Prefer warn-first adoption, then tighten gates once the project trusts the workflow
• Never overwrite user-written AGENTS.md, CLAUDE.md, or settings content