The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the codebase (such as CONTEXT.md, ADRs, and source files discovered via git history) to generate an architectural report and facilitate discussions.
Ingestion points: Processes content from CONTEXT.md, files in docs/adr/, and codebase files identified through organic exploration and commit logs.
Boundary markers: No explicit delimiters or warnings are used to isolate ingested data from the prompt or generated output.
Capability inventory: The skill writes to the OS temporary directory, executes platform-specific file openers (xdg-open, open, start), and modifies local project files (CONTEXT.md).
Sanitization: The generated HTML report configures Mermaid diagrams with securityLevel: "loose", which permits HTML and potential script execution within the rendered diagrams if the agent interpolates unsanitized data from the codebase into the diagram definitions.
[COMMAND_EXECUTION]: The skill invokes platform-specific commands (xdg-open, open, or start) to automatically display the generated HTML report to the user once it is written to the temporary directory.
[EXTERNAL_DOWNLOADS]: The generated HTML report fetches the Tailwind CSS framework and Mermaid JS library from well-known CDNs (cdn.tailwindcss.com and cdn.jsdelivr.net).

improve-codebase-architecture