Orthogonal Agent Framework

v1.0.0

正交性Agent框架 - 模块化AI Agent设计系统。将Agent能力拆分为规划、执行、记忆、评估四个正交维度,实现可组合、可替换、可测试、可扩展的Agent架构。

0· 311·1 current·1 all-time
by李知止@lixingxun
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name/description (orthogonal modular agent framework) match the files and templates provided. Required binary (bash) is minimal and reasonable for templating or filesystem operations described. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md describes creating/maintaining local files (plan.md, execution.md, memory.md, evaluation.md) and using templates — this is appropriate for the framework. There are no instructions to read unrelated system files, access environment secrets, or send data to external endpoints. Note: the framework implies persistent local storage of 'memory' which could capture sensitive user data if used that way; the skill does not provide retention/erase/encryption guidance.
Install Mechanism
No install spec and no code to download or execute. Instruction-only distribution is lowest-risk and matches the content (just templates and docs).
Credentials
No required environment variables, credentials, or config paths. The absence of secrets is proportionate to a documentation/template-only framework.
Persistence & Privilege
always: false and the skill is user-invocable (normal). The framework expects to create and persist template files in the agent/workspace; this is expected but means local data (including any sensitive memory entries) may be stored—users should be aware of where files are written and how long they are retained.
Assessment
This skill is an instruction-only template for structuring agent work (planning, execution, memory, evaluation). It does not request credentials or download code, which reduces risk. Before using it: (1) verify where the agent will write the generated files (workspace path) and whether those files may contain sensitive information; (2) if you will store secrets or private data in the 'memory' files, ensure you have retention/erasure policies or encryption in place; (3) because the source/homepage are not provided, consider testing the skill in a sandboxed environment first and review templates to confirm they meet your operational and privacy requirements.

Like a lobster shell, security has layers — review code before you run it.

latestvk97c0jz7chv0ez2nwhw94ctmth828j9a

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Runtime requirements

🧩 Clawdis
Binsbash

Comments