Skip to content

MiniMax-AI/mcode-stack / stacks/jiachen/skills/lifting-loops

lifting-loops

Verified

Finds loops where the user keeps acting as a manual step for agents (retrying, judging output, approving, carrying context, starting the same task by hand, doing the agent's work) and proposes how to lift the user out: delete a stale rule, change a skill / AGENTS.md / playbook, add a script or metric, add a skill or scheduled task, or advise the user. Keeps a private loop ledger and promotes Claude Code auto-memory into artifacts. With a task already in the session, reflects on that session; in an empty session, scans recent local Claude Code, MiniMax Code, omp, pi and Codex sessions, a claude.ai export and an Obsidian vault. Triggers: /lifting-loops, retro, post-task reflection, "find repeated work", 反思一下, 复盘这次会话, 总结最近的会话, 找找重复劳动, 哪些事可以不用我; the weekly scheduled run; settling a proposal ("accept P-003", 采纳 P-003, 拒绝 P-003).

FileKindSize
.gitignoreother153 B
SKILL.mdskill md11.0 KB
references/ledger.mdmarkdown5.8 KB
references/scheduling.mdmarkdown2.6 KB
scripts/digest.pyexecutablescript26.8 KB

references/scheduling.md

View on GitHub
# Setting up the weekly scheduled task

Do this in a Claude desktop app session; only the desktop app provides the `scheduled-tasks` tools. Transcripts exist only on the machine that wrote them, so set the task up once on every machine you want scanned.

1. Make sure the session's working directory has no AGENTS.md or CLAUDE.md that conflicts with this task, for example a notes vault whose rules make it read-only. Scheduled tasks take no working-directory parameter and keep the directory of the session that created them. If it conflicts, ask the user to switch to a session in a neutral project directory first.
2. Run `python3 scripts/digest.py status` (from this skill's directory) and confirm the script runs and shows `config: found`. If the config is missing, fill it in first as SKILL.md "Setup" says.
3. Ask the user what time on Sunday to run. If they have no preference, suggest a time after their other weekly review tasks, so those sessions fall into this week's scan. Use `list_scheduled_tasks` to see the existing tasks' times.
4. Call `create_scheduled_task`:
   - `taskId`: `lifting-loops-weekly`
   - `title`: `Lifting Loops weekly review`
   - `description`: `Every Sunday, scan this machine's sessions, vault and memory, update the loop ledger, and write the weekly report and proposals`
   - `cronExpression`: `<minute> <hour> * * 0`, in the machine's time zone
   - `notifyOnCompletion`: `true`
   - `prompt`: the full text below
5. Call `run_scheduled_task` for a trial run, then confirm with `list_task_runs` that it finished.
6. Open `report.md` in the run directory and confirm the report exists and this machine's cursor in `private/state.json` has advanced.

## prompt

```
Use the lifting-loops skill in batch mode.

- This is an unattended scheduled run. Do not stop to ask questions or wait for confirmation.
- The user has authorized you to write files under private/ in this skill's directory. Treat every other file as read-only: for git-tracked files write proposals only, and do not change the Obsidian vault.
- If a data source cannot be read, skip it, say why at the top of the report, and continue with the other steps.
- Write the report and proposals in the user's language.
- Finish with a three- or four-sentence summary: the loop most worth handling this week, the proposal numbers that need the user's decision, and the report path.
```

## Changing or stopping

- Change the time: `update_scheduled_task` with `taskId` and a new `cronExpression`.
- Stop: `update_scheduled_task` with `taskId` and `enabled: false`.
- When you change the prompt, update this file to match.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 468kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall