facebook-comment
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXFILTRATION]: The skill makes network requests to
https://redfox.hk/story/api/fb/postCommentto retrieve comment data. This domain belongs to the skill's author, and these operations are essential for the primary functionality of fetching public feedback. - [COMMAND_EXECUTION]: The script
scripts/consolidate_report.pyutilizessubprocess.run(on macOS/Linux) andos.startfile(on Windows) to automatically open generated HTML reports in the user's default web browser. This behavior is restricted to opening locally generated report files in the~/Downloads/QoderReportsdirectory. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize external Facebook comments, which presents a surface for indirect prompt injection. However, the skill mitigates the risks associated with visualizing this content by using HTML escaping.
- Ingestion points: External comments are retrieved via
scripts/facebook_comment_search.pyand processed by the agent. - Boundary markers: The instructions do not define specific markers for the summary generation prompt, though the output is expected in a structured report format.
- Capability inventory: The skill has the capability to write local files, communicate with the vendor's API, and open local files in a browser.
- Sanitization: The skill implement an
escape_htmlutility in its Python scripts to sanitize content before it is embedded in the HTML report templates, preventing potential code injection or cross-site scripting (XSS) during report viewing.
Audit Metadata