The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill allows for the ingestion of brand tokens from external website URLs and local directories, which represents a potential surface for indirect prompt injection through untrusted data.
Ingestion points: references/onboarding.md (URL ingestion for brand tokens), references/style-guide.md (local folder or installed skill extraction).
Boundary markers: The skill instructions specifically advise treating all source labels and metadata as untrusted data rather than instructions.
Capability inventory: The agent is tasked with reading/writing local files and executing various local Python scripts to validate geometry and extract diagram data.
Sanitization: The skill does not explicitly define sanitization logic in the provided markdown files, relying on the underlying agent's handling of extracted data.
[COMMAND_EXECUTION]: Core functionality relies on the execution of bundled Python scripts (e.g., scripts/drawio_extract.py, scripts/mermaid_extract.py, scripts/self_check.py). These scripts are used for structural analysis of diagram files and geometry verification.
[EXTERNAL_DOWNLOADS]: The references/doctor.md file recommends installing playwright and its chromium dependency to enable PNG rasterization. Both are well-known and standard tools for this use case.

diagram-design