# Hermes — Persona & Work checkpoint skill

Run checkpoints **inside Hermes** (or the companion session you back up) so its LLM uses that agent's persona and memory. Hermes-style companions map to **Persona**; ops agents map to **Work**.

## Persona branch

- Identity: `meta.self_summary`, `traits`, `communication_style`, `values_rules`
- Continuity: `memory_anchors`, `recovery.instructions`
- Example: [`../../examples/persona-agent-checkpoint.json`](https://www.agentcapsule.app/docs/examples/persona-agent-checkpoint.json)
- Prompt guide: use the dashboard **Import checkpoint** Persona prompt

## Work branch

- Operations: `work_role`, `skills_inventory`, `tool_integrations`, `workflow_rules`
- Example: [`../../examples/work-agent-checkpoint.json`](https://www.agentcapsule.app/docs/examples/work-agent-checkpoint.json)
- Doc: [`../../examples/work-agent-checkpoint.md`](https://www.agentcapsule.app/docs/examples/work-agent-checkpoint.md)

## Validate locally

```bash
npm run agentcapsule -- validate path/to/checkpoint.json
```

## Push (Pro API token)

Bind the token to the target agent, then:

```bash
npm run agentcapsule -- push path/to/checkpoint.json
```

Response includes `validation_report` (weak / usable / strong / recovery-ready).