The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project files to generate architecture diagrams and dependency reports. A malicious project could contain crafted strings in code comments or manifest files intended to influence the agent's reasoning when it reviews the analysis results.
Ingestion points: The scripts scripts/architecture_diagram_generator.py, scripts/dependency_analyzer.py, and scripts/project_architect.py read source code and configuration files from the user-provided project directory.
Boundary markers: Analysis outputs are generated without explicit delimiters or safety instructions to distinguish analyzed data from the agent's core instructions.
Capability inventory: The skill is limited to local file system reads and standard output; it does not perform network operations, execute arbitrary system commands, or write to sensitive system paths.
Sanitization: Content extraction is performed via regular expressions which filter for specific patterns (like imports), but no sanitization is applied to the extracted text to prevent potential instruction injection.

senior-architect