The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill employs sudo apt-get update and sudo apt-get install -y -qq libreoffice-writer to install system-level software. This grants the skill elevated permissions to modify the underlying operating system environment beyond the user's workspace.
[COMMAND_EXECUTION]: The skill relies heavily on shell command execution via bash to run installers, document converters (pandoc), and office suites (soffice). It also performs runtime modification of the PATH environment variable to include dynamically located binaries.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs multiple Python libraries (pypandoc_binary, typst, openpyxl, python-docx) using pip install --break-system-packages. This flag specifically overrides safety mechanisms (PEP 668) intended to protect the system's Python environment from modification.
[DYNAMIC_EXECUTION]: The skill generates and executes Python logic at runtime through shell heredocs (`python3
<<'PY') and inline script execution (python3 -c`). This facilitates the execution of code constructed dynamically within the skill's flow.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied files, creating a vulnerability surface where malicious instructions within those files could influence agent behavior.
Ingestion points: The skill explicitly processes user-uploaded Word documents (theirs.docx) by converting them to Markdown for editing (found in SKILL.md).
Boundary markers: None. The instructions do not include delimiters or warnings for the agent to ignore instructions embedded within the user's document content.
Capability inventory: The skill possesses capabilities for arbitrary shell execution, Python script execution, and file system write access.
Sanitization: No sanitization or filtering is applied to the content extracted from user documents before it is presented to the agent's context.

office-files