The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bundled Python script, scripts/deskreview.py, to automate Word document manipulation, cell mapping, and evidence extraction. This script performs file system operations, including reading, copying, and writing to .docx and other file types based on user and data inputs.
[EXTERNAL_DOWNLOADS]: The skill instructions recommend installing standard third-party libraries pypdf, python-docx, and openpyxl from official registries to support document processing and text extraction capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources such as Member Association evidence folders containing policies and self-assessments.
Ingestion points: Untrusted data enters the agent's context through text extracted from PDF, DOCX, and XLSX files via the scripts/deskreview.py extract command, which saves the content to a local directory for the agent to read.
Boundary markers: The extraction script adds basic headers (e.g., FILE:, STATUS:) to the output files, but the instructions do not implement strict delimiters or explicit warnings for the agent to ignore instructions embedded in the documents.
Capability inventory: The agent has the ability to read and write local files via the bundled script, allowing it to modify Word documents based on findings from the processed evidence.
Sanitization: There is no evidence of sanitization or filtering of the extracted text content before it is read by the agent.

accreditation-desk-review