The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). Malicious instructions within a document or URL processed by the skill could potentially influence the LLM's summary output.\n
Ingestion points: to-markdown.mjs ingests untrusted data from URLs or local files specified as command-line arguments.\n
Boundary markers: The script uses --- BEGIN DOCUMENT (Markdown) --- and --- END DOCUMENT --- delimiters to isolate document content, though these can be bypassed by identical markers in the input.\n
Capability inventory: The skill performs file system writes and executes shell commands via spawnSync (uvx, pi) in to-markdown.mjs.\n
Sanitization: No sanitization is performed on the input text or user-provided prompts before interpolation into the LLM system prompt.\n- [EXTERNAL_DOWNLOADS]: The skill uses uvx to fetch and run the markitdown utility from the Python Package Index.\n- [COMMAND_EXECUTION]: The script to-markdown.mjs uses spawnSync to execute external commands uvx and pi for document conversion and summarization.

summarize