The Agent Skills Directory

[SAFE]: The skill performs its primary function using specialized MCP tools for project state management and artifact retrieval. No malicious network activity, data exfiltration, or obfuscation was detected.
[COMMAND_EXECUTION]: The skill instructs the user to run a local initialization command (python -m scripts.init_project <slug>) when a project is uninitialized. This is a standard administrative setup procedure and does not represent arbitrary or malicious execution by the agent.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface (Category 8). Untrusted data enters the agent context through the script-os-state.get_project_status and load_artifact tools. While the output is structured using specific markdown headers as boundary markers, no explicit sanitization of artifact content is performed. However, the risk is minimal as the skill's capabilities are restricted to metadata analysis and recommendation rather than generative content creation.

workflow-advisor