The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires installing @coldtea/pr-lens-schema and @coldtea/pr-lens-renderer as direct dependencies. These are required components for the diagram rendering functionality.
[COMMAND_EXECUTION]: The skill executes node for rendering and suggests using rsvg-convert for image conversion. These commands are standard for the intended design illustration workflow.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the user's workspace (requirements, ADRs, source code) to produce diagrams. This presents an attack surface for data-driven influence on the agent's output.
Ingestion points: Reads files such as GLOSSARY.md, ADR.md, and application source code via instructions in SKILL.md.
Boundary markers: None explicitly defined in instructions, but it relies on structured schema-based validation for the rendering data.
Capability inventory: File system writes via writeFile, directory creation via mkdir, and command execution via Node.js and shell.
Sanitization: Employs safeParseGraphDoc from @coldtea/pr-lens-schema for input validation and includes automated tests in test_render.mjs that explicitly check for the absence of script tags in generated SVGs.

illustrate-design