The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to read local HTML files and upload their contents to a user-provided API endpoint. A verification step in the script ensures that only files containing HTML structure tags are processed, which serves as a mitigation against the accidental exfiltration of system configuration or secret files.
[EXTERNAL_DOWNLOADS]: The scripts/publish_html.py script makes network requests to external URLs provided via configuration or command-line arguments.
[COMMAND_EXECUTION]: The skill instructions direct the AI agent to execute local Python scripts and shell commands to facilitate the publishing process.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to the ingestion of external file content. 1. Ingestion points: The html_file argument in scripts/publish_html.py reads data from the local filesystem. 2. Boundary markers: The SKILL.md instructions require the agent to generate or verify the HTML structure, acting as a functional delimiter. 3. Capability inventory: The skill has file system read access and network POST/PUT capabilities. 4. Sanitization: The script validates that the input is a complete HTML document but does not perform content-level sanitization of the HTML payload itself.

publish