The Agent Skills Directory

[SAFE]: The skill is designed for security analysis and documentation. It performs legitimate operations such as reading local repository files and writing a threat model report to the file system. No evidence of malicious activity, data exfiltration, or unauthorized command execution was found.
[COMMAND_EXECUTION]: The skill provides instructions to utilize the rg (Ripgrep) command-line tool with the -I flag to search and explore the codebase. This is a standard and expected function for a tool designed for source code analysis.
[PROMPT_INJECTION]: The skill involves a surface for indirect prompt injection as it processes untrusted source code and configurations from repository paths to generate security documentation. This is an inherent part of its functionality as a threat analysis tool.
Ingestion points: The skill reads source code, architectural components, and data flow evidence from the repository path provided at runtime.
Boundary markers: The internal prompt templates do not explicitly wrap untrusted repository content with delimiters or specific instructions to disregard embedded commands.
Capability inventory: The skill uses file system read access (via rg) and file system write access (to save the final report).
Sanitization: No specific sanitization or filtering logic is described for the code being analyzed.

security-threat-model