The Agent Skills Directory

[SAFE]: The skill package consists entirely of Markdown-based documentation and instructional guidelines. It contains no executable scripts (Python, JavaScript), binary files, or automated logic that could be abused for malicious purposes.
[SAFE]: While the instructions mention sensitive file paths such as .env, ~/.ssh/id_rsa, and /etc/shadow, these are used strictly as pedagogical examples of assets to be identified during an audit or as part of a threat model's asset inventory. There is no evidence of the skill attempting to access or exfiltrate these files from the environment.
[PROMPT_INJECTION]: The skill includes instructions to identify and mitigate prompt injection and 'excessive agency' within the systems being modeled. It does not contain instructions that attempt to bypass the AI agent's own safety guardrails or override its core behavioral constraints.
[COMMAND_EXECUTION]: The skill provides example shell commands (e.g., grep -rn "@app.route") to assist a human or agent in identifying entry points during a security audit. These are presented as static documentation for manual workflow steps and are not configured for automated or dynamic execution with user input.

sota-threat-modeling