Atlas 技術版 > job-registry-maintenance

job-registry-maintenance NO_AGENT UNKNOWN risk=LOW

1. 做什麼 / 2. 為什麼存在

(purpose 尚未填寫 — manifest 為 draft/INFERRED)

3. Schedule / Mode

mode=NO_AGENT state=scheduled enabled=1
schedule=0 4 * * *
next_run=2026-10-22T04:00:00 last_run=2026-10-20T04:01:37
deliver=telegram model=— provider=—

4. Implementation

script=/Users/exion/.hermes/scripts/job_registry_maintenance.py
monitor_script=— monitor_url=—
skills=[]
workdir=—

5. Effective instruction sources

prompt_sha256=85f2b8d07eb37add… chars=861
skill_files={}

This job is no_agent (deterministic script): job_registry_maintenance.py IS the job - it runs with zero LLM and its stdout is delivered verbatim to Telegram. Empty stdout means silent (no message).

Delivery contract - forward each marker verbatim; never rename, merge or reinterpret them:
- [JOB-CONSOLE] Job Console notice/alert (publish notice with fixed URL https://hermes-job-console.pages.dev, or publish-failure notice).
- [RELIABILITY-MAP] Reliability Map notice/alert (external Cloudflare Pages resource health, map publish notice/failure). Always reported separately from Job Console, never as a generic "工作地圖" wording.
- [JOB-REGISTRY] HIGH/CRITICAL drift alert block.
- [GOVERNANCE] / [MAP-STAGE] / [REGISTRY-ERROR] / [REENTRY-GUARD] internal diagnostics.

Rules:
- Never modify production jobs, schedules, or files outside ~/.hermes/job-registry/.
…

6. Actual rules / thresholds

rulevaluesource
REAL/Users/exion/.hermes/job-registry/maintenance.pyCODE_STATIC @job_registry_maintenance.py:10

7/8. Inputs & Outputs (resources)

kindpathnameaccess
none

9/10. Upstream / Downstream

Upstream: —
Downstream: —
Transitive downstream (0): —

Edges (typed evidence)

typehard?fromtoresourceoriginconfevidence
none

11. Shared resources

12. Recent executions

failure_streak=0
last_error=—
delivery_error=—

14. Drift history

16. Impact / change risk

change_risk=LOW