The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes content from untrusted PDF documents.
Ingestion points: Data enters through scripts like scripts/extract_form_field_info.py and scripts/fill_fillable_fields.py using pypdf, and through extraction examples in documentation.
Boundary markers: No explicit delimiters or guardrail instructions are used to distinguish untrusted PDF content from agent instructions.
Capability inventory: The agent can execute local Python scripts via subprocess and perform file system operations to read and write document data.
Sanitization: No sanitization or validation of extracted text is implemented before the agent interprets the data.
[COMMAND_EXECUTION]: The skill's operational logic in forms.md directs the agent to execute shell commands to run internal Python scripts for file conversion, field extraction, and PDF filling.
[SAFE]: The skill relies on well-known and reputable PDF libraries such as pypdf, pdfplumber, and reportlab. The monkeypatch in scripts/fill_fillable_fields.py is a static bug fix for a specific library issue and does not represent a malicious dynamic execution vector.

pdf