| name | loops-playbook |
| description | Choose the simplest loop primitive (turn-based, /goal, /loop, /schedule) and make success criteria explicit and verifiable. |
Loops playbook
This skill summarizes practical guidance for progressing from one-off, turn-based work to goal-based, time-based, and proactive loops, with a focus on clear stop conditions and verifiable success criteria.
Instructions
1) Start with the simplest loop
- Prefer a one-off, turn-based interaction for short tasks that aren’t part of a regular process.
- Escalate to more structured loops only when it materially improves reliability or automation.
2) Make stop conditions explicit
- For goal-based work, define success criteria and an explicit turn cap (e.g., “stop after 5 tries”).
- Avoid subjective success criteria; prefer deterministic checks (tests, scores, thresholds).
3) Improve reliability with verification
- Give Claude a way to verify results end-to-end (tools/connectors, measurable checks).
- When possible, encode manual verification as reusable skills (see
references/related-skill-verify-frontend-change.md).
4) Use the right primitive
- Turn-based: user-triggered, stops when the task is done or more context is needed.
- /goal: user-triggered; stops when the goal is achieved or the turn cap is reached.
- /loop: time-triggered; runs on your computer and stops if you turn it off.
- /schedule: time-triggered in the cloud; suitable for recurring routines.
5) Manage tokens intentionally
- Choose the right primitive/model for the job.
- Pilot before large runs.
- Use scripts for deterministic work.
- Don’t run routines more often than needed.
Examples
See examples/commands.md for command examples from the source post.
References
See references/notes.md for additional notes captured from the source.
Source
