The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the codebase, including source files, ADRs (Architecture Decision Records), and git history, to generate its reports.
Ingestion points: git log --oneline, CONTEXT.md, docs/adr/, and various source files are read into the agent context.
Boundary markers: The prompt does not specify strict boundary markers or instructions to ignore embedded commands within the analyzed code files.
Capability inventory: The agent can write files to the local system (CONTEXT.md, temp HTML reports) and execute shell commands (xdg-open, open, start).
Sanitization: There is no explicit mention of sanitizing codebase content before it is processed or included in the generated HTML report.
[EXTERNAL_DOWNLOADS]: The skill generates an HTML report that loads dependencies from external content delivery networks.
Evidence: The report scaffold in HTML-REPORT.md includes https://cdn.tailwindcss.com and https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.esm.min.mjs.
[COMMAND_EXECUTION]: The skill uses system commands to automatically open the generated report in the user's default browser.
Evidence: Uses xdg-open <path> on Linux, open <path> on macOS, and start <path> on Windows within SKILL.md.
[DYNAMIC_EXECUTION]: The HTML report configuration utilizes a relaxed security posture for its diagramming engine.
Evidence: HTML-REPORT.md initializes Mermaid with securityLevel: "loose", which enables more features but potentially increases exposure to XSS if malicious content from the codebase is included in a diagram.

improve-codebase-architecture