Distributed Job Safety Skill for Idempotent Retries
Keep a stable job identity through duplicate submission and restart, then verify that only one valid output remains.
中文Install in Claude Code
Use your local queue, synthetic input, and a task-state table. The example uses a deterministic job_key.
- Get the complete `plugins/devops-tools/skills/distributed-job-safety` directory from the pinned GitHub commit.
- Copy it into the target project's `.claude/skills/`. Check `.claude/skills/distributed-job-safety/SKILL.md`, start Claude Code at the project root, and call `/distributed-job-safety`.
Example task
Submit fixture-A:v1 twice and simulate one interrupted attempt before retry. Check that state uses one key and that the final output is complete.What to check
The retry reuses the same job identity, producing one valid output with traceable requests.
Original SKILL.md on GitHub · Updated
Frequently asked question
Is checking for an existing job before enqueueing enough?
Concurrent submissions can both pass the check. Use a unique constraint, atomic claim, or equivalent mechanism.