首页 › 龙虾技能列表 › OpenClaw Agent Optimize — OpenClaw 智能体优化

OpenClaw Agent Optimize — OpenClaw 智能体优化

v1.0.0

OpenClaw 智能体优化 1.2.1 工具。

0· 192·4 当前·4 累计
by @zhouhuihui008·MIT-0
下载技能包
License
MIT-0
最后更新
2026/3/25
安全扫描
VirusTotal
无害
查看报告
OpenClaw
安全
high confidence
This is an advisory, instruction-only optimization skill whose requested footprint (no creds, no installs, no code execution in-skill) matches its description — it produces audit/plans and explicitly requires user approval before making persistent changes.
评估建议
This skill is advisory and internally consistent with its stated goal. Before installing or acting on its recommendations: 1) review any exact config/cron patches it proposes and run them in a test/isolated session first; 2) backup your workspace (memory/, cron configs, and skill files) so you can rollback; 3) verify diffs and rollback steps the skill produces before applying changes; 4) if you run the suggested `npx clawhub install` command, confirm the ClawHub package/source you’re installing;...
详细分析 ▾
✓ 用途与能力
The name/description promise a workspace audit and optimization plan for OpenClaw; the skill is purely documentation/instructions and only references editing OpenClaw config, cron patterns, and memory files — all coherent with its stated purpose.
ℹ 指令范围
SKILL.md and the reference docs are advisory and repeatedly instruct not to apply persistent changes without explicit approval. They do reference reading/writing workspace files (memory/*.json, memory/*.md, memory/delegation_log.json) and propose exact config patches (e.g., disabling native heartbeat) — these file and config targets are expected for an optimizer, but any user should verify proposed diffs and backups before applying them.
ℹ 安装机制
There is no built-in install spec (instruction-only, lowest risk). README suggests installing via `npx clawhub install ...` or manual copying; that external npx install is outside the skill bundle and would pull code from ClawHub/npm at runtime — validate that source before running npx. The skill itself does not embed downloads or arbitrary install scripts.
✓ 凭证需求
The skill requests no environment variables, no credentials, and no config paths beyond workspace-local files (memory/, references/, cron definitions). There are no disproportionate secret or credential requests.
ℹ 持久化与权限
always:false (not forced), and model invocation is allowed (normal). The skill can generate config/cron change proposals and exact patches; this is appropriate for its role but means you should not grant the agent permission to apply changes automatically without review. Autonomous invocation combined with user approvals could lead to persistent changes — the docs explicitly require approval before application.
安全有层次，运行前请审查代码。

License

MIT-0

可自由使用、修改和再分发，无需署名。

运行时依赖

无特殊依赖

版本

latestv1.0.02026/3/25

- Added detailed guidance and workflow for optimizing OpenClaw agent workspaces focused on cost, context discipline, routing, and reliability. - Introduced a structured, advisory-first output: audit, options, recommended plan, exact proposal, rollback, and verification steps. - Outlined safety contracts to prevent unauthorized or unsafe changes. - Included quickstart commands, real-world optimization levers, and concise best practices. - Referenced supporting documentation for targeted optimization scenarios.

● 无害

安装命令 点击复制

官方npx clawhub@latest install openclaw-agent-optimize-1-2-1
镜像加速npx clawhub@latest install openclaw-agent-optimize-1-2-1 --registry https://cn.clawhub-mirror.com

技能文档

Use this skill to tune an OpenClaw workspace for cost-aware routing, parallel-first delegation, and lean context.

默认 posture

This skill is advisory first. It should produce:

  • audit,
  • options,
  • recommended plan,
  • exact patch proposal,
  • rollback,
  • verification plan.

No persistent mutations without explicit approval.

Quick 开始

1) Full audit (safe, no changes)

Audit my OpenClaw setup for cost, reliability, and context bloat. Output a prioritized plan with rollback notes. Do NOT apply changes.

2) Context bloat / transcript noise

My OpenClaw context is bloating (slow replies / high cost / lots of transcript noise). Identify the top offenders (tools, crons, bootstrap files, skills) and propose the smallest reversible fixes first. Do NOT apply changes.

3) Model routing / delegation posture

Propose a model routing plan for (a) coding/engineering, (b) short notifications/reminders, (c) reasoning-heavy research/writing. Include an exact config patch + rollback plan, but do NOT apply changes.

什么 good 输出 looks 点赞

  • Executive summary
  • Top drivers
- cost - context - reliability - operator friction
  • Options /B/C 带有 tradeoffs
  • Recommended plan (smallest safe 更改 第一个)
  • Exact proposals + rollback + 验证

Safety contract

  • 做 不 mutate persistent settings 没有 explicit approval.
  • 做 不 创建/更新/移除 cron jobs 没有 explicit approval.
  • 如果 optimization reduces monitoring coverage, present options 和 require choice.
  • 之前 任何 approved 更改, show:
1. exact change, 2. expected impact, 3. rollback plan, 4. post-change verification.

High-ROI optimization levers

1) 输出 discipline 对于 automation

Make maintenance loops truly silent on success.

2) Separate work 从 通知

If you want alerts but want interactive context lean:
  • 做 work quietly
  • notify out-的-band 带有 short human receipt

3) Bootstrap discipline

Keep always-injected files short and load-bearing only. Move long runbooks into references/ or adjacent notes.

4) Ambient specialist surface reduction

A common hidden tax is too many always-visible specialist skills. If a workflow is low-frequency or specialist:
  • prefer 在...上-demand 工作者/subagent usage,
  • 做 不 keep permanently ambient 在...中 main-chat prompt surface.

5) Measure optimizations authoritatively

Prefer fresh-session /context json or equivalent receipts over “feels better”. High-signal fields include:
  • eligible skills
  • skills.promptChars
  • projectContextChars
  • systemPrompt.chars
  • promptTokens

6) Verification-第一个 ops hygiene

After any approved optimization, verify:
  • core chat 仍然 works
  • recall/behavior 做过 不 degrade
  • 新的 会话 actually picks up 更改
  • rollback path proven, 不 theoretical

Workflow (concise)

  • Audit rules + memory: keep restart-critical facts 仅.
  • Audit skill surface: trim ambient specialists 之前 touching tool surface.
  • Audit transcripts/noise: silence cron 和 heartbeat 成功 paths.
  • Audit 模型 routing 和 delegation posture.
  • Recommend smallest viable 更改 第一个.
  • 验证 在...上 新的 会话 当...时 skill/bootstrap snapshotting exists.

Notes

  • 一些 runtimes snapshot skills/配置 per 会话. 如果 您 install/更新 skills 和 做 不 see changes, 开始 新的 会话.
  • Prefer short SKILL.md + references/ 对于 long runbooks.
  • 如果 context bloat main complaint, pair skill 带有 context-clean-up (audit-仅).

References

  • references/optimization-playbook.md
  • references/模型-selection.md
  • references/context-management.md
  • references/agent-orchestration.md
  • references/cron-optimization.md
  • references/heartbeat-optimization.md
  • references/memory-patterns.md
  • references/continuous-learning.md
  • references/safeguards.md
数据来源：ClawHub ↗ · 中文优化：龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制

免费技能或插件可能存在安全风险，如需更匹配、更安全的方案，建议联系付费定制

了解定制服务