The Agent Skills Directory

[SAFE]: The skill enforces a secure workflow by requiring explicit user confirmation through AskUserQuestion before generating any agent files, preventing unauthorized file system modifications.
[SAFE]: Project analysis (L2) is conducted using standard search and read tools to understand local context, with no evidence of data exfiltration or sensitive information exposure.
[SAFE]: File generation is handled by a local script (scripts/init_agent.cjs) that uses a static template and basic string replacement. The script contains no dynamic execution sinks such as eval() or exec().
[SAFE]: The skill documentation explicitly instructs the agent to enforce project-relative boundaries and reject absolute or global paths, minimizing the risk of directory traversal or workspace escape.
[SAFE]: No obfuscation, malicious prompt injection patterns, or unverifiable remote dependencies were detected across the skill's files.

project-agent-writer