The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill generates an HTML report that initializes the Mermaid library with securityLevel: "loose". This configuration allows for the execution of HTML tags and scripts embedded within diagrams. Since the skill populates these diagrams using data scanned from the local codebase, a malicious file in the repository could lead to Cross-Site Scripting (XSS) in the browser when the user opens the report.
[COMMAND_EXECUTION]: The skill automatically executes platform-specific commands (xdg-open on Linux, open on macOS, and start on Windows) to launch the generated HTML report in the user's default web browser.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to the way it processes external data. 1. Ingestion points: The skill reads codebase files, commit history (git log), CONTEXT.md, and ADRs in docs/adr/. 2. Boundary markers: None identified. The skill does not use specific delimiters or instructions to ignore embedded commands in the scanned data. 3. Capability inventory: The skill can write files to the system temporary directory, execute shell commands to open browsers, and call sub-agents. 4. Sanitization: No sanitization or escaping mechanisms are described for the data interpolated into the HTML report or sub-agent context.
[EXTERNAL_DOWNLOADS]: The generated report includes script tags that fetch resources from cdn.tailwindcss.com and cdn.jsdelivr.net. These are well-known and established content delivery networks used for styling and rendering the visual report.

improve-codebase-architecture