← 返回 Skills 市场
Tokenoptimizer
作者
smartpeopleconnected
· GitHub ↗
· v1.0.18
11086
总下载
27
收藏
102
当前安装
19
版本数
在 OpenClaw 中安装
/install token-optimizer
功能描述
Reduce OpenClaw AI costs by 97%. Haiku model routing, free Ollama heartbeats, prompt caching, and budget controls. Go from $1,500/month to $50/month in 5 min...
安全使用建议
Install only if you want a tool that can change OpenClaw cost settings. Run preview commands first, inspect generated diffs and prompt files, and use --apply only after confirming the model routing, budgets, heartbeat provider, memory rules, and local stats/report files are acceptable.
功能分析
Type: OpenClaw Skill
Name: token-optimizer
Version: 1.0.18
The OpenClaw Token Optimizer skill bundle is benign. Its code and documentation are entirely focused on optimizing OpenClaw AI costs through model routing, heartbeat configuration, caching, and budget controls. The `CHANGELOG.md` explicitly details significant security improvements, such as the removal of all `subprocess.run` calls in favor of safer `shutil.which` and `urllib.request` for health checks, and defaulting commands to dry-run mode. Instructions for the AI agent in `SKILL.md` and template files (`templates/SOUL.md`, `templates/OPTIMIZATION-RULES.md`) are defensive, guiding the agent towards cost-aware and responsible behavior, rather than attempting prompt injection. There is no evidence of data exfiltration, malicious execution, persistence, or obfuscation.
能力评估
Purpose & Capability
The capabilities fit the stated purpose: model routing, heartbeat configuration, caching, budgets, rollback, verification, and prompt templates are all tied to reducing OpenClaw token costs.
Instruction Scope
The triggers are broad and the generated prompt rules affect future agent behavior, but the behavior is visible, cost-focused, and not automatically injected without user action.
Install Mechanism
The package uses a local Python CLI with no runtime dependencies or post-install execution shown; state-changing optimization commands default to preview unless --apply is used.
Credentials
File access is mostly scoped to ~/.openclaw plus a local analysis report, and network activity is limited to provider reachability checks such as localhost Ollama/LM Studio or Groq; no credential harvesting or exfiltration path was found.
Persistence & Privilege
It can persist config, backups, workspace templates, prompt rules, stats, and an analysis report; these are mostly disclosed, though analyze/verify and dry-run directory creation are minor least-surprise gaps.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install token-optimizer - 安装完成后，直接呼叫该 Skill 的名称或使用
/token-optimizer触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.18
Remove slack target, default all CLI commands to dry-run, add file modification docs
v1.0.17
- Removed 27 files, including all marketing materials, PDFs, scripts, and comparison skills to streamline the codebase.
- Cleaned up documentation in SKILL.md and README.md for a more focused and concise overview.
- Dropped references to removed features and scripts from user guides.
- No core logic changes to the skill; optimization features remain unchanged.
v1.0.16
Version 1.0.16 introduces major new skills, marketing materials, and testing updates.
- Added new skills including agent-memory, better-memory, context-optimizer, personas, and startclaw-optimizer for expanded optimization capabilities.
- Introduced marketing folder with plans and social media post templates to support outreach.
- Added comprehensive PDF documentation and status reports.
- Updated core modules and tests for stability and integration of new features.
- Enhanced documentation and marketplace materials for improved onboarding and visibility.
v1.0.15
Docs consistency: removed all references to install scripts, pure Python CLI only
v1.0.14
Exclude install scripts from publish - pure Python only, no shell script execution
v1.0.13
Exclude marketing files from publish to resolve scanner findings
v1.0.12
Security: removed curl|sh pattern, installer now downloads to temp with user confirmation before execution
v1.0.11
Security scanner clean: removed all unicode control chars, sanitized prompt references, softened changelog wording
v1.0.10
Security scan fix: removed emoji unicode chars and prompt-injection patterns from SKILL.md
v1.0.9
Updated listing with v1.0.8 features: multi-provider heartbeat, rollback, health check, diff preview
v1.0.8
v1.0.8: Multi-provider heartbeat (ollama/lmstudio/groq/none), rollback command, health check, diff preview in dry-run, --no-color flag, shared colors module, extended triggers, version and license fixes
v1.0.7
Added discovery tags
v1.0.6
Added viral discovery keywords and agent-friendly documentation
v1.0.5
Added simulation test for performance comparison
v1.0.4
Fixed YAML frontmatter - using description field
v1.0.3
Added summary in YAML frontmatter
v1.0.2
Added summary and updated metadata
v1.0.1
Improved skill description and documentation
v1.0.0
- Initial public release of Token Optimizer.
- Provides configuration analyzer and basic optimization templates.
- Includes prompt caching, model routing, and session management guidelines.
- Added quick start instructions and clear cost-saving breakdowns.
- Documentation includes free and premium feature comparison, installation guides, and generated configuration examples.
元数据
常见问题
Tokenoptimizer 是什么？
Reduce OpenClaw AI costs by 97%. Haiku model routing, free Ollama heartbeats, prompt caching, and budget controls. Go from $1,500/month to $50/month in 5 min... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 11086 次。
如何安装 Tokenoptimizer？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install token-optimizer」即可一键安装，无需额外配置。
Tokenoptimizer 是免费的吗？
是的，Tokenoptimizer 完全免费（开源免费），可自由下载、安装和使用。
Tokenoptimizer 支持哪些平台？
Tokenoptimizer 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Tokenoptimizer？
由 smartpeopleconnected（@smartpeopleconnected）开发并维护，当前版本 v1.0.18。
推荐 Skills