fetch-content
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external URLs, which represents an indirect prompt injection attack surface.
- Ingestion points: External content is fetched from URLs in
scripts/fetch.py. - Boundary markers: The skill implements a safety boundary called the "untrusted-content contract," wrapping fetched data in
<untrusted-content>tags to distinguish it from agent instructions. - Capability inventory: The tool is designed for read-only text normalization and does not have the ability to write files or execute code.
- Sanitization: It uses regular expressions to neutralize attempts to close the safety fence and strips control characters from the output to prevent terminal escapes or hidden content.
- [EXTERNAL_DOWNLOADS]: The script fetches data from well-known services including YouTube, TikTok, and Twitter/X (via
fxtwitter.com) to provide transcripts and engagement metadata.
Audit Metadata