The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell scripts to perform diagram conversion and version control operations.
Evidence: The convert-drawio-to-png.sh script executes the drawio CLI for image export and git add to automatically stage generated assets.
Evidence: The skill instructions in SKILL.md suggest using mise exec and pre-commit for workflow automation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted XML data from .drawio files, which presents a surface where embedded malicious instructions could influence agent behavior.
Ingestion points: The agent is directed to read, edit, and parse .drawio XML files as part of its core functionality (SKILL.md, README.md).
Boundary markers: The instructions lack specific boundary markers or "ignore instructions" warnings when processing the contents of diagram files.
Capability inventory: The skill has the ability to invoke subprocesses, including the drawio binary, the git client, and a local Python search utility (find_aws_icon.py).
Sanitization: There is no evidence of sanitization or strict schema validation of the XML text content prior to processing by the agent.

draw-io