Token Cost Optimizer
v1.0.0OpenClaw Token 成本优化与 API 用量控制专家。当用户提到 token 消耗过高、API 费用上涨、 模型调用太贵、上下文太长、死循环、定时任务频率、多 Agent 重复请求、模型路由、 或任何与降低 OpenClaw 运行成本相关的话题时，立即使用此 Skill。 也适用于用户说"帮我省钱"、"...
⭐ 0· 127·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description (token cost optimization) matches the instructions: diagnosing cron/jobs, inspecting openclaw config and memory/workspace sizes, changing compaction settings, model routing, and pruning old files. No unrelated credentials, binaries, or external services are requested.
Instruction Scope
SKILL.md instructs the agent to run OpenClaw CLI commands and to read/edit files under ~/.openclaw (openclaw.json, soul.md, memory/workspace). These are within the skill's stated scope, but it also includes destructive actions (find ... -delete, removing unused skills/memory, openclaw gateway stop/start). The instructions grant broad discretion (e.g., '删除长期未用的 Skill 目录') without explicit safeguards or dry-run steps.
Install Mechanism
Instruction-only skill with no install spec, no downloads, and no code files—lowest install risk.
Credentials
No environment variables, credentials, or config paths outside ~/.openclaw are requested. The requested filesystem access is proportional to modifying OpenClaw configuration and cleaning OpenClaw workspaces.
Persistence & Privilege
The skill is not always-on and does not request persistent elevated privileges or modify other skills' configs. It advises editing OpenClaw files but does not itself install components or assert permanent presence.
Assessment
This skill appears to do what it says (optimize token usage) and makes sensible recommendations, but several instructions are potentially disruptive: they delete files and restart gateways. Before using it, (1) ensure you have recent backups of ~/.openclaw (configs, memory, workspace), (2) run the diagnostic commands manually to inspect outputs, not automatically, (3) replace destructive commands with safe dry-runs first (e.g., run find ... -print before -delete), (4) verify the referenced model names and routing match your environment, and (5) apply changes in a staging environment if possible. Because the skill is instruction-only, the agent will execute any CLI commands you allow — review each command and confirm you trust the OpenClaw CLI and the machine running these commands.Like a lobster shell, security has layers — review code before you run it.
latestvk970mp082rc29ng0076pdv3vks83cecz
License
MIT-0
Free to use, modify, and redistribute. No attribution required.