The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains Python functions that execute external binaries using the subprocess module to render diagrams and manage integrations.
Evidence: subprocess.run([engine, ...]) in the render_graphviz function (SKILL.md) is used to invoke Graphviz layout engines.
Evidence: subprocess.run(["joplin", "import", ...]) in the publish_graphviz_to_joplin function (SKILL.md) interacts with the Joplin CLI for data import.
[INDIRECT_PROMPT_INJECTION]: The skill processes Graphviz DOT source code, which serves as an ingestion surface for potentially untrusted data that could influence agent behavior if not handled carefully.
Ingestion points: The dot_source parameter in render_graphviz, publish_graphviz_to_obsidian, and publish_graphviz_to_joplin functions (SKILL.md).
Boundary markers: No explicit boundary markers or delimiters for the untrusted DOT source are defined in the provided code templates.
Capability inventory: The skill utilizes subprocess.run for command execution and pathlib.Path for file system writes, including directory creation (mkdir) and file writing (write_text).
Sanitization: There is no evidence of sanitization, validation, or escaping of the dot_source input before it is passed to the rendering engine or written to the file system.
[PRIVILEGE_ESCALATION]: The documentation includes instructions for users or agents to acquire administrative privileges to install dependencies.
Evidence: The Prerequisites section in SKILL.md explicitly lists sudo apt-get install graphviz as a requirement for Linux environments.

graphviz-diagrams