The Agent Skills Directory

[SAFE]: The skill implements a 'Deterministic Security Layer' via the verify_anchors.py script. This script automatically scans model-generated HTML for malicious patterns, including <script> tags, <iframe> tags, event handlers (e.g., onclick), and external resource references (e.g., https:// URLs in src or href).
[SAFE]: Data integrity is enforced through a 'Lexical Anchor' and 'Number Fidelity' contract. The verification scripts ensure that any content or numbers presented in the final visual reader are strictly grounded in the original source Markdown, preventing the AI from hallucinating or following hidden instructions in the input data.
[SAFE]: Input handling is secure. The md_source.py script, which renders the authoritative source view, uses standard HTML escaping to prevent Cross-Site Scripting (XSS) from malicious Markdown input.
[SAFE]: The skill operates locally on the user's machine. The provided Python and Node.js scripts use standard libraries for file processing and do not perform unexpected network operations or access sensitive system paths.
[SAFE]: The execution pipeline requires a human-in-the-loop 'Self-Review' step (screenshot verification), providing an additional layer of oversight for the final output.

md2view