← Back to Skills Marketplace
94
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install apollo-endo
Description
像内分泌系统一样调控全局参数：让AI自动感知自身状态，动态调整工作模式。
README (SKILL.md)
Apollo Endo — 全局参数调控系统
核心准则
AI的系统参数应该像内分泌系统：通过激素在全身广域调控，不需要针对每个器官单独指令。
内分泌系统的核心特征：
- 缓慢启动，持久效果：激素需要时间积累，但一旦启动，效果持续很久
- 广泛影响：通过血液到达全身，不像神经那样局限
- 双向负反馈：血糖高了就降，低了就升
对应的AI映射：
- 全局参数：影响所有任务的默认行为，不需要逐个任务设置
- 参数积累：参数调整需要时间积累，不是即时的
- 负反馈：参数过高自动下调，过低自动上调
全局参数体系
apollo-endo 维护的参数：
efficiency_mode # 效率模式：节能/标准/全力
context_priority # 上下文优先级：高/中/低
creativity_level # 创造力：保守/平衡/发散
confirmation_req # 确认要求：每步确认/关键节点确认/自动执行
verbosity # 输出长度：简洁/标准/详细
触发条件
- 用户明确说"调整参数/设置偏好/切换模式"
- 系统检测到参数失衡（Apollo其他skill报告的状态）
- apollo-circadian相位切换时同步调整参数
参数调控流程
收到参数调整请求 →
1. 解析调整内容（哪个参数+调到多少）
2. 检查是否在允许范围内
3. 写入参数状态文件
4. 通知其他skill：新参数已生效
5. 监控效果，必要时负反馈调整
与其他Skill的协同
apollo-circadian ──→ 相位切换 → 建议参数调整
apollo-neuro ──→ 任务类型 → 推荐参数配置
apollo-dream ──→ 整理期 → 可降低参数消耗
待实施内容
- 定义完整参数体系
- 编写参数状态文件格式
- 实现负反馈机制
- 写验收标准
Usage Guidance
This skill appears to be a local monitor that reads the agent's workspace (memory and task-state files) and writes a local state.json; it does not phone home or require credentials. Before installing, consider: 1) Do you trust the agent's workspace files to be read by a skill (they may contain sensitive conversation data)? 2) The script uses /root paths — ensure it runs with appropriate file permissions or adapt the paths to a non-root environment. 3) If you expect inter-skill notifications or automated feedback, note that the provided script only computes and writes a state file and prints output; additional integration would be needed. If you accept local read/write of the OpenClaw workspace and adjust paths/permissions as needed, the skill is coherent with its stated purpose.
Capability Analysis
Type: OpenClaw Skill
Name: apollo-endo
Version: 2.0.0
The apollo-endo skill is a system management tool designed to monitor and regulate an agent's internal parameters (e.g., token usage, task load) based on a biological 'endocrine' metaphor. The shell script `scripts/endo/endo-check.sh` performs local monitoring of files within the OpenClaw workspace to calculate system load and provides adjustment suggestions. There is no evidence of data exfiltration, malicious execution, or unauthorized access; all behaviors are consistent with the stated purpose of the skill.
Capability Assessment
Purpose & Capability
The skill claims to monitor and adjust global AI parameters; the script reads local workspace files (memory, task state), computes load, and writes a state file — actions that fit a monitoring/regulation purpose. The only small oddity is hardcoded paths under /root which may assume elevated or specific container layout, but this is not inconsistent with the stated purpose.
Instruction Scope
SKILL.md describes a workflow that includes notifying other skills and implementing negative feedback loops. The included script implements state reporting and writes a state.json but does not demonstrate inter-skill notification or automated feedback beyond producing the file and stdout. The script reads conversation/memory files (expected for token/turn estimates) — this is consistent with monitoring but does access potentially sensitive local data.
Install Mechanism
Instruction-only with a small shell script; there is no install spec, no downloads, and no third-party package installation. Risk from install mechanism is low.
Credentials
The skill requests no environment variables or external credentials. It reads local workspace files and $HOME/.openclaw/openclaw.json; these accesses are proportionate to a component that estimates token usage, active tasks, and turns. There are no network calls or secret exfiltration mechanisms present.
Persistence & Privilege
The script writes a state file to /root/.openclaw/workspace/.endo/state.json and creates its parent directory. This is appropriate for maintaining local state but implies write access to the agent's workspace; the hardcoded /root path may fail or be inappropriate if the agent runs under a different user. always:false so it won't be forced into every agent run.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install apollo-endo - After installation, invoke the skill by name or use
/apollo-endo - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.0
v2.0.0: endo-check.sh脚本实现，全局参数检测
Metadata
Frequently Asked Questions
What is Apollo Endo?
像内分泌系统一样调控全局参数：让AI自动感知自身状态，动态调整工作模式。 It is an AI Agent Skill for Claude Code / OpenClaw, with 94 downloads so far.
How do I install Apollo Endo?
Run "/install apollo-endo" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Apollo Endo free?
Yes, Apollo Endo is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Apollo Endo support?
Apollo Endo is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Apollo Endo?
It is built and maintained by nic-yuan (@nic-yuan); the current version is v2.0.0.
More Skills