The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted markdown files which may contain malicious instructions or payloads. This risk is mitigated through design and sanitization.\n
Ingestion points: The scripts scripts/diff_parser.py and scripts/annotation_extractor.py read user-provided markdown files from the local filesystem.\n
Boundary markers: None.\n
Capability inventory: The skill performs file reading and writing within the user-specified project directories.\n
Sanitization: All potentially untrusted content, such as annotation bodies and PR titles, is sanitized via html.escape() in scripts/review_html_renderer.py before being included in the generated HTML.\n- [EXTERNAL_DOWNLOADS]: The skill's HTML renderer references external typography resources from Google Fonts to maintain visual style.\n
Evidence: Reference to https://fonts.googleapis.com/css2 found in scripts/review_html_renderer.py and the associated template file.\n- [SAFE]: The application code consists of local Python scripts using only the standard library. There is no evidence of obfuscation, hardcoded credentials, or unauthorized network activity.

md-review