# Agent Persistence

> Enable and explain the reusable human-in-the-loop gates used by persistent agent loops for destructive actions, false-positive overrides, and recovery escalation.

- Skill: `jmagly/agent-persistence` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jmagly/agent-persistence`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jmagly/agent-persistence/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: jmagly (https://skillmd.com/u/jmagly)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jmagly/agent-persistence

---


# Agent Persistence

Use this driver to activate or understand the `agent-persistence` addon:

```bash
aiwg use agent-persistence
```

The addon supplies deterministic gate definitions, not a replacement loop
runtime. Compose it with `agent-loop` or another orchestrator that supports
human checkpoints.

## Canonical gates

- `destructive-action-gate.yaml` requires authorization before destructive work.
- `false-positive-override-gate.yaml` records reviewed policy overrides.
- `recovery-escalation-gate.yaml` escalates bounded recovery failures.

Read the gate contract before integrating it, preserve its authorization
boundary, and use the owning loop's status surface to verify activation.

## References

- @$AIWG_ROOT/agentic/code/addons/agent-persistence/docs/hitl-integration.md
- @$AIWG_ROOT/agentic/code/addons/agent-loop/skills/agent-loop/SKILL.md