The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from project files to generate structured models, which represents an indirect prompt injection surface.\n
Ingestion points: The skill reads requirements from features.md, business_rules.md, and requirements_analysis_report.md as specified in the Input Files section of SKILL.md.\n
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between its own logic and potentially malicious instructions embedded in the source files.\n
Capability inventory: The skill is restricted to file system operations (read and write) on Markdown artifacts. It does not perform network requests, spawn subprocesses, or use dynamic evaluation tools.\n
Sanitization: No evidence of data sanitization or validation of the input text was found in the core instructions.\n- [SAFE]: No high-risk malicious patterns were detected. The skill does not contain hardcoded credentials, remote code downloads, persistence mechanisms, or obfuscated content. Its behavior is consistent with its stated purpose of assisting in technical specification authoring.

06-requirements-patterns