The Agent Skills Directory

[COMMAND_EXECUTION]: The setup.sh script executes shell commands to install system dependencies and Python packages. This is used to set up the environment with required tools like LibreOffice and libraries such as python-docx and pyyaml.
[PRIVILEGE_ESCALATION]: The setup.sh script uses sudo to install the LibreOffice suite on Linux systems. The skill instructions advise the agent to ask for user permission before installation, mitigating the risk of unauthorized privilege escalation.
[EXTERNAL_DOWNLOADS]: The skill fetches and installs software packages from well-known registries, including PyPI (via pip) and official operating system package managers (e.g., apt, brew, dnf, pacman). These downloads are directed to trusted organizations and services for the purpose of document conversion and manipulation.
[INDIRECT_PROMPT_INJECTION]: The skill processes external Word documents (.doc, .docx) and YAML configuration files, which represents an attack surface for indirect prompt injection. If an agent reads a malicious template using inspect_template.py, it could be exposed to hidden instructions embedded in the document text.
Ingestion points: scripts/inspect_template.py and scripts/fill_form.py read content from user-provided files.
Boundary markers: The skill does not implement explicit markers to distinguish between data and instructions in the output of the inspection scripts.
Capability inventory: The skill has file-writing capabilities (scripts/fill_form.py) and command execution/installation capabilities (setup.sh).
Sanitization: No sanitization is performed on the text read from the documents before it is presented to the agent.

fill-protocol