The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill automates the setup of its execution environment by downloading the official .NET SDK installer.
The scripts/docx bash script fetches the installation script from https://dot.net/v1/dotnet-install.sh, which is a well-known service provided by Microsoft.
[COMMAND_EXECUTION]: The skill relies on executing system commands and pre-compiled binaries to perform its core functions.
The scripts/docx utility compiles and executes C# code (dotnet build and dotnet run) generated by the agent to create documents.
Pre-compiled binaries and DLLs in the validator/ directory (e.g., Validator.dll) are executed to validate OpenXML compliance.
The scripts/md2docx/md2docx_convert.py script invokes pandoc via subprocess.run to convert Markdown to Word format.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the way it processes external data.
Ingestion points: Untrusted data enters the agent context through Markdown files and citation databases (citation.jsonl) in scripts/md2docx/md2docx_convert.py, and via existing Word documents in scripts/engine/.
Boundary markers: The skill does not use explicit delimiters or instructions to ignore embedded commands when processing these external files.
Capability inventory: The skill possesses significant capabilities, including arbitrary file writing and shell command execution via the .NET runtime and pandoc.
Sanitization: Data processing involves standard JSON and XML parsing, but does not include content-level sanitization to strip potential malicious instructions before they are processed by the agent.

docx