| name | skillopt-sleep-kimi |
| version | 1.0.0 |
| description | Use when the user wants their Kimi CLI agent to self-improve from past usage, asks about a nightly/offline "sleep" or "dream" cycle, memory/skill consolidation, or says things like "make my agent better the more I use it", "review my past sessions", "learn my preferences", "consolidate what you learned", "run the sleep cycle", or wants to schedule offline self-optimization.
|
| author | devbase-team |
| tags | ["skillopt","self-improvement","memory","kimicli","sleep-cycle"] |
| skill_type | custom |
| inputs | [{"name":"scope","type":"string","description":"Scope of the sleep cycle (current-project | all)","default":"current-project","required":false},{"name":"backend","type":"string","description":"Backend for offline replay (mock | local | remote)","default":"mock","required":false}] |
| outputs | [{"name":"report","type":"markdown","description":"Summary of harvested sessions, mined tasks, and staged proposals"}] |
SkillOpt-Sleep for Kimi CLI: offline self-evolution
This skill gives the user's Kimi CLI agent a sleep cycle. While the user is
offline (or on demand), it reviews real past sessions, re-runs recurring tasks,
and consolidates what it learns into memory and skills — but only keeps
changes that pass a held-out validation gate, and only after the user adopts them.
When to activate
Trigger when the user wants any of:
- "make my agent learn from how I use it" / "get better the more I use it"
- a nightly/scheduled or on-demand offline self-improvement / dream / sleep run
- to review past sessions/trajectories and distill recurring tasks
- to consolidate feedback into devbase Vault notes or managed skills
- to run
status, harvest, dry-run, run, or adopt for SkillOpt-Sleep
The cycle
- Harvest — use
devkit_session_recall and devkit_vault_history to gather
recent sessions and note changes (read-only).
- Mine — turn session digests into recurring
TaskRecords with outcomes and
checkable references where possible.
- Replay — re-run mined tasks offline under the current skill and memory.
- Consolidate — reflect on failures and propose bounded edits to Vault notes
or skills.
- Gate — accept edits only when a held-out validation score improves.
- Stage — write the proposal under
<project>/.skillopt-sleep/staging/<date>/; nothing live changes.
- Adopt — only after explicit user approval, copy staged files over live
files with backups.
How to drive it
Because Kimi CLI does not have a plugin script model like Claude Code, the cycle
is driven through devbase MCP tools and shell commands:
python -m skillopt_sleep status --project "$(pwd)"
python -m skillopt_sleep dry-run --project "$(pwd)" --backend mock
python -m skillopt_sleep run --project "$(pwd)" --backend local
python -m skillopt_sleep adopt --project "$(pwd)"
For context gathering, also use:
devkit_session_recall — find relevant past sessions.
devkit_vault_search — locate existing memory/skills related to the mined tasks.
devkit_knowledge_report — summarize the current project state before proposing edits.
Hard rules
Validate / demo
python -m skillopt_sleep.experiments.run_experiment --persona researcher --assert-improves
python -m skillopt_sleep.experiments.run_experiment --persona programmer --assert-improves
See the SkillOpt-Sleep guide for recorded output and
docs/superpowers/specs/2026-06-07-skillopt-sleep-claude-code-plugin-design.md
for the full design.