The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructions to proactively override the agent's default terminal behavior. It specifically directs the agent to replace ASCII-style tables and diagrams with styled HTML pages automatically when certain complexity thresholds are met.
[EXTERNAL_DOWNLOADS]: Generated HTML files incorporate several well-known and reputable third-party libraries via the JSDelivr CDN. These include Mermaid.js for diagramming, Chart.js, Observable Plot, and D3.js for data visualization, and anime.js for animations. It also references Google Fonts for typography. These are standard, well-known services and do not represent a security risk.
[COMMAND_EXECUTION]: The skill makes legitimate use of system commands and command-line tools to gather information about the project and display results. These include:
Version control tools: git and gh (GitHub CLI) for analyzing code changes, history, and pull requests.
File utilities: ls, grep, wc, and cat for scanning files and aggregating metrics.
Browser integration: open (macOS) and xdg-open (Linux) to launch the generated HTML files.
Optional tools: surf-cli for generating AI-based conceptual illustrations.
[PROMPT_INJECTION]: The skill processes untrusted external data such as git diffs, commit messages, and implementation plans. While this represents an indirect prompt injection surface, the risk is mitigated by the agent's instructions to verify claims against the codebase and the localized, browser-based nature of the output.

visual-explainer