The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes a bulk push script that ingests local markdown files and transmits their contents via API calls.
Ingestion points: The Bulk Push Pattern section in SKILL.md provides a Python script that reads all *.md files from a user-defined directory using glob.glob and open().
Capability inventory: The script uses urllib.request to perform POST operations to a remote API endpoint, effectively exporting the content of local files to an external service.
Boundary markers: The script does not utilize boundary markers or instructions to the agent to ignore or isolate potentially malicious instructions that could be embedded within the processed markdown files.
Sanitization: While the script performs regex transformations to convert markdown tables into lists, it lacks sanitization mechanisms to filter or escape prompt injection patterns or sensitive data before transmission.

fluent-community