The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the codebase, specifically project files, CONTEXT.md, and architectural decision records (ADRs), to generate its findings.
Ingestion points: Project source code and documentation files (CONTEXT.md, docs/adr/*.md).
Boundary markers: None specified for separating codebase content from agent instructions during the exploration phase.
Capability inventory: File system reading, writing to temporary directories, system command execution to open files, and the ability to trigger downstream skills (/grilling, /domain-modeling).
Sanitization: There is no evidence of sanitization for strings extracted from the codebase before they are embedded into the HTML report.
[DYNAMIC_EXECUTION]: The skill generates a self-contained HTML report at runtime. It configures the Mermaid visualization library with securityLevel: "loose". This configuration allows the rendering of HTML tags within diagram labels, which increases the risk of Cross-Site Scripting (XSS) if the agent includes malicious strings from the codebase into the diagram's source text.
[COMMAND_EXECUTION]: The skill uses platform-specific system commands (xdg-open, open, start) to automatically launch the generated report in the user's web browser.
[EXTERNAL_DOWNLOADS]: The generated report fetches external libraries from well-known services at runtime, including Tailwind CSS from cdn.tailwindcss.com and the Mermaid library from cdn.jsdelivr.net.

improve-codebase-architecture