The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external documents such as PDF, Word, and images to generate answers and reports.\n
Ingestion points: The scripts/generate_report.py script reads files from a user-provided directory.\n
Boundary markers: The skill does not use specific delimiters for document content in the agent prompt, but instead instructs the agent to treat the parsed JSON as context.\n
Capability inventory: The skill can write JSON files to the system temporary directory, write HTML reports to the local project directory, and open the local report in a web browser.\n
Sanitization: The report generation script uses html_module.escape to sanitize text extracted from documents before embedding it into the HTML template, mitigating XSS risks.\n- [EXTERNAL_DOWNLOADS]: The skill requires the liteparse Python package and @llamaindex/liteparse Node package. These originate from LlamaIndex, which is a well-known service in the AI development ecosystem.\n- [COMMAND_EXECUTION]: The skill uses a bundled Python script to manage the parsing pipeline and report generation. The execution of this script is the primary function of the skill and is restricted via platform configuration.

research-docs