The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses shell commands to invoke local helper scripts and system utilities for document processing.
Evidence includes execution of scripts/office/soffice.py, scripts/office/unpack.py, scripts/office/pack.py, scripts/office/validate.py, scripts/accept_changes.py, and scripts/comment.py.
It also utilizes system binaries such as pandoc and pdftoppm for text extraction and image conversion.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted .docx files which serves as an attack surface for indirect prompt injection.
Ingestion points: Document content is ingested into the agent context via pandoc extraction or XML unpacking of .docx files (referenced in SKILL.md).
Boundary markers: The instructions do not define specific delimiters or security warnings to isolate extracted content from the agent's instructions.
Capability inventory: The skill possesses capabilities for file writing (via pack.py and JavaScript fs.writeFileSync) and command execution (via listed local scripts).
Sanitization: No evidence of content sanitization or filtering of extracted document text is provided in the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external software to function.
It instructs the user to install the docx package globally via npm install -g docx.
It references system-level dependencies including pandoc, LibreOffice, and Poppler.
[DYNAMIC_EXECUTION]: The skill generates and executes JavaScript code at runtime to create document buffers.
The agent is instructed to write JavaScript logic using the docx library and execute it to produce file output via Packer.toBuffer and fs.writeFileSync.

docx