The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documents provided by users, which serves as an ingestion point for potentially malicious instructions. This creates an indirect prompt injection surface where the agent's behavior could be influenced by content within the analyzed files.
Ingestion points: Ingests data from PDF, DOCX, PPTX, and XLSX files.
Boundary markers: The instructions do not specify the use of strict boundary markers when interpolating document text into subsequent prompt steps.
Capability inventory: Utilizes web_search, bash (for calculations), and local Python scripts for document modification.
Sanitization: The skill identifies and flags non-public information, but does not explicitly describe sanitization of inputs before processing.
[COMMAND_EXECUTION]: The skill uses the bash tool to execute Python one-liners (python -c) for performing numerical consistency checks and runs several local scripts (annotate_pdf.py, annotate_pptx.py, annotate_xlsx.py) to modify documents.
[EXTERNAL_DOWNLOADS]: The workflow incorporates the web_search tool during the fact-checking phase to verify document claims against external public data sources.

document-review