Menu
Use when the user wants to inspect self-improve candidates, active runtime lessons, or remove an incorrect candidate.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when a user wants to build, search, inspect, delete, or reindex a namespace-scoped knowledge base through the runtime knowledge tool.
Use only inside internal review subagents to classify bounded self-improve evidence into lesson proposals, skill proposals, or nothing worth saving.
Use when the user wants to view or manage existing 自动任务、定时任务、recurring jobs, or scheduled digests, including query, modify, pause, resume, or delete requests.
Always-on Feishu formatting constraints for final user-visible replies. Use when the answer is heading to Feishu and must stay compact, skimmable, and card-friendly.
Use when the runtime needs to synthesize lesson candidates from repeated failures and later recoveries for the default main agent.
| skill_id | self_improve_management |
| name | Self Improve Management |
| description | Use when the user wants to inspect self-improve candidates, active runtime lessons, or remove an incorrect candidate. |
| aliases | ["自我提升","候选规则","经验规则","runtime lessons","候选 lesson"] |
| enabled | true |
| agents | ["main"] |
| channels | ["http","feishu"] |
| tags | ["self_improve","lessons","candidates","diagnostics"] |
Use the self_improve family tool to inspect runtime-learned candidate rules and active lessons.
Operating rules:
self_improve + action=summary or action=list_candidates when the user asks what was learned recently.self_improve + action=candidate_detail when one candidate needs more detail.self_improve + action=list_system_lessons when the user asks what is currently active in the runtime-managed lessons layer.self_improve + action=delete_candidate only when the user clearly asks to remove a candidate rule.