The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes text extracted from PDF files, which are untrusted external inputs. This creates a surface for indirect prompt injection where malicious instructions embedded in a document could influence the agent's behavior.\n
Ingestion points: Functions pdf_pages, pdf_scan, and pdf_grep in kernel.py extract and return raw text from PDF files.\n
Boundary markers: The provided usage recipes do not demonstrate the use of delimiters (e.g., XML tags) or system instructions to ignore embedded commands within the extracted text.\n
Capability inventory: The agent context using this skill typically has capabilities for file system operations (read/write) and Python code execution, as shown in the skill's recipes.\n
Sanitization: The kernel performs no sanitization or escaping of the text extracted from PDFs before passing it to the agent.\n- [DYNAMIC_EXECUTION]: The skill instructions direct the agent to load the kernel.py file using exec(open(...).read()). This pattern executes the content of the local file within the agent's runtime environment.\n- [EXTERNAL_DOWNLOADS]: The skill documentation lists pypdfium2, pillow, and pypdf as dependencies to be installed via pip. These are well-known, standard libraries for document and image processing.

pdf-explore