← Back to Skills Marketplace
perrykono-debug

Agent Efficiency Manager

by perrykono-debug · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
35
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install agent-efficiency-manager
Description
定期分析所有 Agent 的 token 使用效率,智能推荐技能配置,生成优化报告并自动推送,实现 Agent 长期自我优化。
README (SKILL.md)

SKILL.md

License: MIT
Copyright: 2026 perrykono-debug



name: agent-efficiency-manager description: Agent 效率优化与自我进化管理器。定期分析所有 Agent 的 token 使用效率,识别可优化的技能配置,从 skillhub 发现并推荐新技能,自动推送优化建议,长期跟踪效率指标实现自我进化。触发场景:(1) 用户要求"优化 agent"、"降低 token 成本"、"提升效率", (2) 定期效率分析与推送, (3) 发现新技能并评估适用性, (4) 生成效率报告与优化建议

Agent Efficiency Manager

Agent 效率优化与自我进化管理器,通过定期分析、智能推荐和持续跟踪,实现 Agent 长期自我优化。

Workflow Decision Tree

用户请求优化
    ├─ 分析模式 → Run analyze_agent_efficiency.py
    ├─ 推荐模式 → Run fetch_skillhub_skills.py + generate_recommendations.py
    ├─ 跟踪模式 → Run track_metrics.py
    └─ 推送模式 → Run push_notifications.py

Core Capabilities

1. 效率分析(analyze_agent_efficiency.py)

扫描 openclaw.json 中所有 Agent 配置,计算效率指标:

指标定义(参见 references/metrics_definitions.md):

  • skill_count:技能数量
  • estimated_tokens:预估 token 消耗(每个 skill 约 500-1000 tokens)
  • efficiency_score:效率评分(0-100,基于 skill 相关度)
  • redundancy_count:重复/无关技能数量

执行:

python3 scripts/analyze_agent_efficiency.py --config ~/.qclaw/openclaw.json --output metrics.json

输出: JSON 格式效率报告,包含每个 Agent 的详细指标和优化建议。


2. 技能推荐(fetch_skillhub_skills.py + generate_recommendations.py)

从 skillhub 获取可用技能列表,基于 Agent 角色推荐适配技能:

执行:

# 步骤1:获取 skillhub 技能列表
python3 scripts/fetch_skillhub_skills.py --output available_skills.json

# 步骤2:生成推荐
python3 scripts/generate_recommendations.py --config ~/.qclaw/openclaw.json --available available_skills.json --output recommendations.json

推荐逻辑(参见 references/recommendation_logic.md):

  • 匹配 Agent 角色与技能描述
  • 排除已安装的技能
  • 优先推荐高评分、高频使用的技能
  • 检测技能冲突(功能重复)

3. 指标跟踪(track_metrics.py)

长期记录效率指标,生成趋势分析:

执行:

python3 scripts/track_metrics.py --config ~/.qclaw/openclaw.json --history metrics_history.json --output trend_report.md

跟踪内容:

  • Token 消耗趋势(按周/月)
  • 技能数量变化
  • 效率评分演变
  • 优化建议采纳率

4. 自动推送(push_notifications.py)

通过企微/腾讯文档推送优化建议:

执行:

python3 scripts/push_notifications.py --recommendations recommendations.json --channel wecom --webhook YOUR_WEBHOOK_URL

推送内容:

  • 每周效率报告
  • 新技能推荐(附使用场景说明)
  • 配置优化建议(具体 skills 增删列表)
  • 成本节省预估

Quick Start

场景 1:全量效率分析

用户说:"分析所有 Agent 的效率,给出优化建议"

执行流程:

  1. 运行 analyze_agent_efficiency.py 生成效率报告
  2. 运行 fetch_skillhub_skills.py 获取可用技能
  3. 运行 generate_recommendations.py 生成推荐
  4. 输出综合优化建议(Markdown 格式)

场景 2:定期监控与推送

用户说:"每周一早上 9 点推送效率报告"

执行流程:

  1. 创建 cron 任务(qclaw-cron-skill
  2. 定时运行分析脚本
  3. 生成报告并推送(企微/webchat)

场景 3:技能发现与评估

用户说:"有没有新技能可以提升 Stock 大作手的效果?"

执行流程:

  1. 运行 fetch_skillhub_skills.py
  2. 针对 "stock" Agent 运行 generate_recommendations.py
  3. 输出推荐技能列表(含评分和理由)

Optimization Patterns

常见优化模式(详见 references/optimization_patterns.md):

模式 做法 Token 节省
去重 移除 another_them / another-them 重复 ~1000
去无关 移除跨领域技能(股票 Agent 不留房产技能) ~500-2000
精简基础 只保留 1-2 个基础配置技能 ~500-1000
按需启用 一次性工具(如 qclaw-migration)不常驻 ~500

Resources

scripts/

  • analyze_agent_efficiency.py:效率分析脚本
  • fetch_skillhub_skills.py:获取 skillhub 技能列表
  • generate_recommendations.py:生成优化建议
  • track_metrics.py:跟踪长期指标
  • push_notifications.py:推送通知

references/

  • metrics_definitions.md:效率指标定义与计算方法
  • optimization_patterns.md:常见优化模式与案例
  • recommendation_logic.md:技能推荐逻辑详解

assets/

(本技能无需静态资源)


Notes

  • 预估 token 消耗:每个 skill 的 SKILL.md 约 500-1000 tokens,实际消耗取决于模型和内容长度
  • 安全操作:所有优化建议默认为"建议"模式,需用户确认后才执行 config.patch
  • 定期运行:建议每周运行一次完整分析,每日运行快速检查
  • 自我进化:跟踪优化建议的采纳情况,持续改进推荐算法
Usage Guidance
Install only if you are comfortable with the skill reading your OpenClaw agent configuration. Keep reports local unless you intentionally provide a trusted webhook, because pushed reports may reveal agent names, installed skills, and optimization recommendations.
Capability Assessment
Purpose & Capability
The stated purpose is agent efficiency analysis, skill recommendation, metrics tracking, and optional report delivery; the scripts match that purpose by reading an OpenClaw config, producing local JSON/Markdown reports, and optionally sending a summarized report to a user-provided webhook.
Instruction Scope
The skill describes periodic analysis and push workflows, but the artifact does not create schedules by itself; scheduling is framed as a user-requested workflow. It also states optimization changes are recommendations and require user confirmation before config.patch.
Install Mechanism
The package contains markdown documentation, reference files, and Python helper scripts. There is no install-time hook, obfuscation, dependency installer, or automatic persistence mechanism in the artifacts.
Credentials
Reading ~/.qclaw/openclaw.json and listing all agents is proportionate to an efficiency manager, but users should expect agent names, skill lists, and recommendation details to be included in generated reports.
Persistence & Privilege
The documentation suggests using a cron skill for weekly reports and supports webhook delivery, but no script independently installs cron jobs or modifies agent configuration. Webhook use requires an explicit destination parameter.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install agent-efficiency-manager
  3. After installation, invoke the skill by name or use /agent-efficiency-manager
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Agent Efficiency Manager 1.0.0 – Initial Release - Introduces an automated agent efficiency optimization and self-improvement manager. - Provides scripts for periodic efficiency analysis, skill recommendations, metrics tracking, and automated report notifications. - Defines core metric calculations: skill count, estimated token usage, efficiency score, and redundancy. - Integrates with skillhub for new skill discovery and relevance-based recommendations. - Supports scheduled monitoring, trend tracking, and detailed markdown or JSON report generation.
Metadata
Slug agent-efficiency-manager
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Agent Efficiency Manager?

定期分析所有 Agent 的 token 使用效率,智能推荐技能配置,生成优化报告并自动推送,实现 Agent 长期自我优化。 It is an AI Agent Skill for Claude Code / OpenClaw, with 35 downloads so far.

How do I install Agent Efficiency Manager?

Run "/install agent-efficiency-manager" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Agent Efficiency Manager free?

Yes, Agent Efficiency Manager is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Agent Efficiency Manager support?

Agent Efficiency Manager is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Agent Efficiency Manager?

It is built and maintained by perrykono-debug (@perrykono-debug); the current version is v1.0.0.

💬 Comments