The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill generates an HTML report that references styling and diagramming libraries from well-known services.
Evidence: The HTML-REPORT.md file specifies the use of the Tailwind CSS CDN and the JSDelivr CDN for Mermaid.js in the report scaffold.
[COMMAND_EXECUTION]: The skill uses standard operating system commands to display the generated architectural report to the user.
Evidence: Uses xdg-open (Linux), open (macOS), or start (Windows) to launch the HTML file created in the temporary directory.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its ingestion of untrusted codebase data to drive its analysis and report generation.
Ingestion points: The agent is instructed to read CONTEXT.md, Architecture Decision Records (ADRs) in docs/adr/, and general project source code using an exploration sub-agent.
Boundary markers: There are no explicit instructions to use delimiters or to ignore potential instructions embedded within the ingested codebase files.
Capability inventory: The skill has permissions to write to CONTEXT.md and ADR files, execute shell commands to open reports, and invoke other agent skills such as /grilling, /domain-modeling, and /codebase-design.
Sanitization: No sanitization or validation of the ingested codebase content is described, and the Mermaid configuration in the report uses securityLevel: "loose", which allows for a broader range of HTML and script interactions within diagrams generated from codebase data.

improve-codebase-architecture