The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill suggests using npx @david-choi/devkit and npx @truongnat/devkit to analyze documents. While the latter is a vendor resource, the former originates from an external third-party developer.\n- [REMOTE_CODE_EXECUTION]: The use of npx in references/attachment-preprocessing.md allows for the execution of remote code from the npm registry. The skill also suggests optional installation of packages via pip.\n- [COMMAND_EXECUTION]: The skill executes local scripts and binary tools, specifically node .agents/devkit/dist/tools.js and markitdown, to extract text from various document formats.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by reading and processing content from external documents (PDF, DOCX, etc.) provided by the user.\n
Ingestion points: Files are processed in references/attachment-preprocessing.md via the analyze-doc tool.\n
Boundary markers: The skill uses structured sections (Provided files:, Extracted context:) to separate file data from instructions, but these are not foolproof against adversarial content.\n
Capability inventory: The skill has the ability to execute shell commands (node, npx) and install packages.\n
Sanitization: No specific sanitization or instruction-filtering is mentioned for the text extracted from documents.

router-pro