The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references a specification document hosted on the official Google Cloud Platform GitHub repository and the generated graph viewer (viz.html) fetches legitimate libraries (cytoscape and marked) from the jsDelivr CDN to enable graph visualization and markdown rendering.\n- [INDIRECT_PROMPT_INJECTION]: The scripts/render_okf_viewer.py script creates a vulnerability surface by processing untrusted markdown files for display in an HTML viewer.\n
Ingestion points: The collect function in scripts/render_okf_viewer.py reads all .md files within the user-provided bundle directory.\n
Boundary markers: The script does not implement boundary markers or instructions to ignore embedded malicious content when processing files.\n
Capability inventory: The script has the capability to write files to the local filesystem (index.md and viz.html).\n
Sanitization: While basic metadata is escaped, the markdown body is passed directly to the marked.parse function in the generated viz.html without sanitization. This could allow an attacker to execute arbitrary JavaScript in the user's browser if they are tricked into rendering a malicious OKF bundle.

okf