[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6MS13ZKM38HJAP1CWW19W0X":3,"skill-same:01M6MS13ZKM38HJAP1CWW19W0X":125},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":22,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":59,"commitSha":60,"skillDirectory":61,"forksCount":13,"repositoryHtmlUrl":62,"repositoryHomepageUrl":6,"ownerHtmlUrl":63,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":64,"install":65,"resources":78},"01M6MS13ZKM38HJAP1CWW19W0X","openclaw-agent-optimize",null,"Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace\n(cost, model routing, context discipline, delegation, reliability).\nDon't use when: you want immediate config\u002Fcron mutations without review, or the question is unrelated to OpenClaw ops.\nOutput: a prioritized plan with exact change proposals, expected impact, and rollback steps. No persistent changes without explicit approval.","urjuyaimon09\u002Fxiage-openclaw-workspace","urjuyaimon09","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F231897024?v=4","01M6MRMWFSYW8EX4SHQHSVGNTT","01M6MRMWKGQVHAWASPF1ZT6DXT",0,"MIT","https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002FSKILL.md",[17],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"openclaw agent optimize","optimize","openclaw agent","openclaw","question","you","impact","options","unrelated","discipline","reliability","context","model","plan","improve","audit","delegation","config","change","review","2026-11-10 10:12:30",false,"---\nname: openclaw-agent-optimize\nslug: openclaw-agent-optimize\nversion: 1.2.1\nlicense: MIT\ndescription: |\n  Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace\n  (cost, model routing, context discipline, delegation, reliability).\n  Don't use when: you want immediate config\u002Fcron mutations without review, or the question is unrelated to OpenClaw ops.\n  Output: a prioritized plan with exact change proposals, expected impact, and rollback steps. No persistent changes without explicit approval.\ntriggers:\n  - optimize agent\n  - optimizing agent\n  - improve OpenClaw setup\n  - agent best practices\n  - OpenClaw optimization\nmetadata:\n  openclaw:\n    emoji: \"?\"\n---\n\n# OpenClaw Agent Optimization\n\nUse this skill to tune an OpenClaw workspace for **cost-aware routing**, **parallel-first delegation**, and **lean context**.\n\n## Default posture\n\nThis skill is **advisory first**.\nIt should produce:\n- audit,\n- options,\n- recommended plan,\n- exact patch proposal,\n- rollback,\n- verification plan.\n\nNo persistent mutations without explicit approval.\n\n## Quick start\n\n1) **Full audit (safe, no changes)**\n> Audit my OpenClaw setup for cost, reliability, and context bloat. Output a prioritized plan with rollback notes. Do NOT apply changes.\n\n2) **Context bloat \u002F transcript noise**\n> My OpenClaw context is bloating (slow replies \u002F high cost \u002F lots of transcript noise). Identify the top offenders (tools, crons, bootstrap files, skills) and propose the smallest reversible fixes first. Do NOT apply changes.\n\n3) **Model routing \u002F delegation posture**\n> Propose a model routing plan for (a) coding\u002Fengineering, (b) short notifications\u002Freminders, (c) reasoning-heavy research\u002Fwriting. Include an exact config patch + rollback plan, but do NOT apply changes.\n\n## What good output looks like\n\n- **Executive summary**\n- **Top drivers**\n  - cost\n  - context\n  - reliability\n  - operator friction\n- **Options A\u002FB\u002FC** with tradeoffs\n- **Recommended plan** (smallest safe change first)\n- **Exact proposals** + **rollback** + **verify**\n\n## Safety contract\n\n- Do not mutate persistent settings without explicit approval.\n- Do not create\u002Fupdate\u002Fremove cron jobs without explicit approval.\n- If an optimization reduces monitoring coverage, present options and require choice.\n- Before any approved change, show:\n  1. exact change,\n  2. expected impact,\n  3. rollback plan,\n  4. post-change verification.\n\n## High-ROI optimization levers\n\n### 1) Output discipline for automation\nMake maintenance loops truly silent on success.\n\n### 2) Separate work from notification\nIf you want alerts but want interactive context lean:\n- do the work quietly\n- notify out-of-band with a short human receipt\n\n### 3) Bootstrap discipline\nKeep always-injected files short and load-bearing only.\nMove long runbooks into `references\u002F` or adjacent notes.\n\n### 4) Ambient specialist surface reduction\nA common hidden tax is **too many always-visible specialist skills**.\nIf a workflow is low-frequency or specialist:\n- prefer on-demand worker\u002Fsubagent usage,\n- do not keep it permanently ambient in main-chat prompt surface.\n\n### 5) Measure optimizations authoritatively\nPrefer fresh-session `\u002Fcontext json` or equivalent receipts over “feels better”.\nHigh-signal fields include:\n- `eligible skills`\n- `skills.promptChars`\n- `projectContextChars`\n- `systemPrompt.chars`\n- `promptTokens`\n\n### 6) Verification-first ops hygiene\nAfter any approved optimization, verify:\n- core chat still works\n- recall\u002Fbehavior did not degrade\n- new session actually picks up the change\n- rollback path is proven, not theoretical\n\n## Workflow (concise)\n\n1. Audit rules + memory: keep restart-critical facts only.\n2. Audit skill surface: trim ambient specialists before touching tool surface.\n3. Audit transcripts\u002Fnoise: silence cron and heartbeat success paths.\n4. Audit model routing and delegation posture.\n5. Recommend the smallest viable change first.\n6. Verify on a **new session** when skill\u002Fbootstrap snapshotting exists.\n\n## Notes\n\n- Some runtimes snapshot skills\u002Fconfig per session. If you install\u002Fupdate skills and do not see changes, start a new session.\n- Prefer short `SKILL.md` + `references\u002F` for long runbooks.\n- If context bloat is the main complaint, pair this skill with `context-clean-up` (audit-only).\n\n## References\n\n- `references\u002Foptimization-playbook.md`\n- `references\u002Fmodel-selection.md`\n- `references\u002Fcontext-management.md`\n- `references\u002Fagent-orchestration.md`\n- `references\u002Fcron-optimization.md`\n- `references\u002Fheartbeat-optimization.md`\n- `references\u002Fmemory-patterns.md`\n- `references\u002Fcontinuous-learning.md`\n- `references\u002Fsafeguards.md`\n","# OpenClaw Agent Optimization\n\nUse this skill to tune an OpenClaw workspace for **cost-aware routing**, **parallel-first delegation**, and **lean context**.\n\n## Default posture\n\nThis skill is **advisory first**.\nIt should produce:\n- audit,\n- options,\n- recommended plan,\n- exact patch proposal,\n- rollback,\n- verification plan.\n\nNo persistent mutations without explicit approval.\n\n## Quick start\n\n1) **Full audit (safe, no changes)**\n> Audit my OpenClaw setup for cost, reliability, and context bloat. Output a prioritized plan with rollback notes. Do NOT apply changes.\n\n2) **Context bloat \u002F transcript noise**\n> My OpenClaw context is bloating (slow replies \u002F high cost \u002F lots of transcript noise). Identify the top offenders (tools, crons, bootstrap files, skills) and propose the smallest reversible fixes first. Do NOT apply changes.\n\n3) **Model routing \u002F delegation posture**\n> Propose a model routing plan for (a) coding\u002Fengineering, (b) short notifications\u002Freminders, (c) reasoning-heavy research\u002Fwriting. Include an exact config patch + rollback plan, but do NOT apply changes.\n\n## What good output looks like\n\n- **Executive summary**\n- **Top drivers**\n  - cost\n  - context\n  - reliability\n  - operator friction\n- **Options A\u002FB\u002FC** with tradeoffs\n- **Recommended plan** (smallest safe change first)\n- **Exact proposals** + **rollback** + **verify**\n\n## Safety contract\n\n- Do not mutate persistent settings without explicit approval.\n- Do not create\u002Fupdate\u002Fremove cron jobs without explicit approval.\n- If an optimization reduces monitoring coverage, present options and require choice.\n- Before any approved change, show:\n  1. exact change,\n  2. expected impact,\n  3. rollback plan,\n  4. post-change verification.\n\n## High-ROI optimization levers\n\n### 1) Output discipline for automation\nMake maintenance loops truly silent on success.\n\n### 2) Separate work from notification\nIf you want alerts but want interactive context lean:\n- do the work quietly\n- notify out-of-band with a short human receipt\n\n### 3) Bootstrap discipline\nKeep always-injected files short and load-bearing only.\nMove long runbooks into `references\u002F` or adjacent notes.\n\n### 4) Ambient specialist surface reduction\nA common hidden tax is **too many always-visible specialist skills**.\nIf a workflow is low-frequency or specialist:\n- prefer on-demand worker\u002Fsubagent usage,\n- do not keep it permanently ambient in main-chat prompt surface.\n\n### 5) Measure optimizations authoritatively\nPrefer fresh-session `\u002Fcontext json` or equivalent receipts over “feels better”.\nHigh-signal fields include:\n- `eligible skills`\n- `skills.promptChars`\n- `projectContextChars`\n- `systemPrompt.chars`\n- `promptTokens`\n\n### 6) Verification-first ops hygiene\nAfter any approved optimization, verify:\n- core chat still works\n- recall\u002Fbehavior did not degrade\n- new session actually picks up the change\n- rollback path is proven, not theoretical\n\n## Workflow (concise)\n\n1. Audit rules + memory: keep restart-critical facts only.\n2. Audit skill surface: trim ambient specialists before touching tool surface.\n3. Audit transcripts\u002Fnoise: silence cron and heartbeat success paths.\n4. Audit model routing and delegation posture.\n5. Recommend the smallest viable change first.\n6. Verify on a **new session** when skill\u002Fbootstrap snapshotting exists.\n\n## Notes\n\n- Some runtimes snapshot skills\u002Fconfig per session. If you install\u002Fupdate skills and do not see changes, start a new session.\n- Prefer short `SKILL.md` + `references\u002F` for long runbooks.\n- If context bloat is the main complaint, pair this skill with `context-clean-up` (audit-only).\n\n## References\n\n- `references\u002Foptimization-playbook.md`\n- `references\u002Fmodel-selection.md`\n- `references\u002Fcontext-management.md`\n- `references\u002Fagent-orchestration.md`\n- `references\u002Fcron-optimization.md`\n- `references\u002Fheartbeat-optimization.md`\n- `references\u002Fmemory-patterns.md`\n- `references\u002Fcontinuous-learning.md`\n- `references\u002Fsafeguards.md`\n",{"name":5,"slug":5,"license":14,"version":48,"metadata":49,"triggers":52,"description":58},"1.2.1",{"openclaw":50},{"emoji":51},"?",[53,54,55,56,57],"optimize agent","optimizing agent","improve OpenClaw setup","agent best practices","OpenClaw optimization","Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace\n(cost, model routing, context discipline, delegation, reliability).\nDon't use when: you want immediate config\u002Fcron mutations without review, or the question is unrelated to OpenClaw ops.\nOutput: a prioritized plan with exact change proposals, expected impact, and rollback steps. No persistent changes without explicit approval.\n","master","ece10d94aa9ed5e21ddc1700c5ec060d19d1295b","skills\u002Fphenomenoner-openclaw-agent-optimize","https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace","https:\u002F\u002Fgithub.com\u002Furjuyaimon09","2026-04-19 17:35:31",{"source":8,"skillName":5,"commands":66,"prompt":76,"zipAvailable":77},[67,70,73],{"tool":68,"command":69},"npx","npx skills add urjuyaimon09\u002Fxiage-openclaw-workspace --skill openclaw-agent-optimize",{"tool":71,"command":72},"bunx","bunx skills add urjuyaimon09\u002Fxiage-openclaw-workspace --skill openclaw-agent-optimize",{"tool":74,"command":75},"pnpm","pnpm dlx skills add urjuyaimon09\u002Fxiage-openclaw-workspace --skill openclaw-agent-optimize","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002FSKILL.md",true,[79,85,89,93,97,101,105,109,113,117,121],{"path":80,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":83,"sourceHtmlUrl":84},"_meta.json","REFERENCE","参考文档",142,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002F_meta.json",{"path":86,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":87,"sourceHtmlUrl":88},"README.md",3004,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002FREADME.md",{"path":90,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":91,"sourceHtmlUrl":92},"references\u002Fagent-orchestration.md",1042,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fagent-orchestration.md",{"path":94,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":95,"sourceHtmlUrl":96},"references\u002Fcontext-management.md",1072,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fcontext-management.md",{"path":98,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":99,"sourceHtmlUrl":100},"references\u002Fcontinuous-learning.md",898,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fcontinuous-learning.md",{"path":102,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":103,"sourceHtmlUrl":104},"references\u002Fcron-optimization.md",3201,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fcron-optimization.md",{"path":106,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":107,"sourceHtmlUrl":108},"references\u002Fheartbeat-optimization.md",6509,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fheartbeat-optimization.md",{"path":110,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":111,"sourceHtmlUrl":112},"references\u002Fmemory-patterns.md",901,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fmemory-patterns.md",{"path":114,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":115,"sourceHtmlUrl":116},"references\u002Fmodel-selection.md",1146,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fmodel-selection.md",{"path":118,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":119,"sourceHtmlUrl":120},"references\u002Foptimization-playbook.md",3026,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Foptimization-playbook.md",{"path":122,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":123,"sourceHtmlUrl":124},"references\u002Fsafeguards.md",2315,"https:\u002F\u002Fgithub.com\u002Furjuyaimon09\u002Fxiage-openclaw-workspace\u002Fblob\u002Fmaster\u002Fskills\u002Fphenomenoner-openclaw-agent-optimize\u002Freferences\u002Fsafeguards.md",[]]