- name
- daemon-cleanup
- description
- Daemon footprint scope, reclaim persistence, and consent-gated cleanup of old artifacts.
- version
- 1.0.2
- last_changed_at
- 2026-09-08T00:00:00Z
- related_files
- ["src/lingtai/tools/skills/manual/reference/cleanup-footprint-contract.md","src/lingtai/tools/daemon/manual/SKILL.md","src/lingtai/tools/daemon/manual/reference/forensics/SKILL.md"]
- maintenance
- Keep this reference aligned with the daemon artifact layout and the shared
footprint recipe; update the owner route when either changes.
# Daemon Cleanup Reference
Use this reference for a footprint audit or explicitly authorized deletion.
It does not own provider routing or cross-process recovery. There is no
automatic folder cleanup: `reclaim` stops processes but leaves evidence, and
molt clears conversation context while retaining durable stores, including
daemon run folders.
## Cleanup / Footprint
Each run leaves `daemons/em-*` with `daemon.json`, events, transcript/history,
result files, and token records. Do not delete an active run or evidence still
needed for a report, review, or cost audit.
Call `psyche(action="skills", input={}, reasoning="read cleanup guidance")`,
then read `reference/cleanup-footprint-contract.md` beside that installed manual.
Use its shared inspection recipe; combine
its definitions with this tool-specific selection in one task-owned script.
Inspection writes nothing. Appending `logs/cleanup.jsonl` is a separate,
explicitly selected audit step.
```python
agent = Path.cwd()
items = [p for p in (agent / "daemons").glob("em-*") if p.is_dir()]
rows, total = footprint_check(items, tool="daemon", top_n=20)
```
Before deleting old completed folders, show the dry-run output and obtain
explicit human consent. Then append an `apply` record to
`logs/cleanup.jsonl` with the deleted paths and bytes. Never delete outside the
approved selection or use cleanup to hide a failed run.
在 GitHub 查看