The Agent Skills Directory

[SAFE]: No security vulnerabilities or malicious patterns detected.
The skill includes an explicit 'Canon Write Gate' logic that prohibits the agent from mutating existing files unless the user provides an unambiguous and explicit directive, effectively mitigating the risk of accidental or malicious data modification via prompt injection.
File operations are restricted to the .lev/pm/ directory, which is defined as a vendor-specific storage path for the author 'lev-os', representing legitimate project management functionality.
The skill protocol requires structured data transformation ('compiling intent') and fidelity scoring before any durable data is written, which adds a verification layer between untrusted user input and stored artifacts.
No network exfiltration, hardcoded credentials, or obfuscated payloads were identified.

capture