The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of several well-known Python packages for document generation and visualization:
reportlab and matplotlib for PDF structure and chart rendering.
pypdf and pymupdf for page count verification and rasterization.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to manage its environment:
Creation of a Python virtual environment (venv) to ensure isolation.
Execution of a local Python script (scripts/render_report.py) to process data and output the PDF file.
[INDIRECT_PROMPT_INJECTION]: The skill serves as a rendering pipeline for findings that may originate from untrusted sources (e.g., automated scans of third-party code).
Ingestion points: The agent writes user-provided or tool-provided findings into a findings.json file, which is consumed by the rendering script.
Boundary markers: The findings are structured according to a defined JSON schema, providing clear boundaries between data fields.
Capability inventory: The skill has file-write access (for the JSON and PDF) and command execution capabilities (for the rendering script and dependency management).
Sanitization: The render_report.py script includes an esc() function utilizing xml.sax.saxutils.escape to sanitize text values before they are passed to ReportLab's Paragraph flowables, mitigating potential injection into the PDF generator.

audit-report