The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands such as cargo run, just build, and cargo clean within the ~/code/diagrams directory to render figures and manage build artifacts.
[DYNAMIC_EXECUTION]: The core functionality relies on generating or modifying Rust source code and then invoking the Rust compiler and runner (cargo) to produce SVG and PNG outputs.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text ('copy') and interpolates it into Rust code constructors (e.g., txt("...")). This creates an attack surface where maliciously crafted user input could attempt to escape string boundaries to execute arbitrary Rust code during the build process.
Ingestion points: User requests for diagram content and text labels (referenced in references/authoring.md and references/overflow.md).
Boundary markers: The instructions do not specify explicit delimiters or sanitization steps for the user-supplied text before interpolation into code.
Capability inventory: Access to the local filesystem to write .rs files and execution of the shell via cargo and just (referenced in references/build.md).
Sanitization: No specific sanitization or escaping logic is described for handling external content in the provided documentation.

svg-diagram