The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard system commands such as git, grep, wc, and ls to analyze the codebase for architectural friction. It also executes platform-specific commands like open, start, and xdg-open to display the generated HTML report to the user. Additionally, it runs a local Python script (scripts/verify-report.py) to validate the report's integrity.
[EXTERNAL_DOWNLOADS]: The generated HTML report references style and diagramming libraries from well-known services (Tailwind CSS and JSDelivr) to render content in the user's browser.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the codebase, including source files, CONTEXT.md, and Architecture Decision Records (ADRs). This represents an attack surface where malicious content within the repository could attempt to influence the agent's analysis.
Ingestion points: Files read during the scoping and exploration phases (e.g., source code, CONTEXT.md, docs/adr/*.md).
Boundary markers: None specified in the instructions for delimiting or ignoring embedded instructions in external file content.
Capability inventory: The skill performs file reading, local command execution (git, grep, python), and opens a browser to view a generated report.
Sanitization: No explicit sanitization, validation, or filtering of file content is described before processing or interpolation.

improve-codebase-architecture