The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from PDF files which is interpolated into prompts for the Anthropic API for semantic extraction.\n
Ingestion points: The scripts/docmap.py utility reads external PDF files provided by the user via the pdfplumber library.\n
Boundary markers: The LLM prompts utilize structured headers such as SECTION, PAGES, and TEXT to delimit the document content from instructions.\n
Capability inventory: The skill communicates with the Anthropic API and writes extracted data files (_MAP.md, .symbols.json, .anchors.json) to the local file system.\n
Sanitization: The script performs JSON validation on the API response and includes error handling for parsing failures.\n- [EXTERNAL_DOWNLOADS]: The skill installs the pdfplumber and anthropic Python packages. The anthropic package is a standard library provided by a well-known and trusted organization.

mapping-documents