| name | agentic-governance |
| description | Portable governance layer for long-running AI coding work. Use when the challenge is keeping a whole program bounded, resumable, and evidence-backed across sessions and agents. |
Agentic Governance
Use this skill when the work is big enough that transcript memory and workflow momentum are no longer safe enough by themselves.
Overview
Agentic Governance is a governance layer, not a replacement execution engine.
It introduces:
- objective maps
- bounded packets
- wave limits
- blocker taxonomy
- evidence logs
- fail-closed closeout
Workflow
- Establish or refresh the
objective-map.json.
- Open a bounded wave in
wave-state.json.
- Create exact packet artifacts under
packets/.
- Choose one adapter for packet execution.
- Require the adapter to report back into governance artifacts.
- Fail closed if scope spills, packets remain open, or evidence is incomplete.
Checklist
- Do the artifacts exist?
- Is there exactly one active wave?
- Does each packet have an owned and forbidden surface?
- Is the chosen adapter allowed for the packet?
- Did execution update evidence and packet history?
