yt-viral
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes YouTube video titles and channel data provided in an external JSON file. If these titles contain malicious instructions, they could influence the agent's final summary or structural analysis.\n
- Ingestion points: The
swipe.pyscript reads video titles and channel data from a user-supplied JSON file (e.g.,collected.json) through theopen(files[0])call.\n - Boundary markers: The instructions in
SKILL.mddefine how to present the results but lack specific instructions or delimiters to isolate untrusted video titles from the agent's reasoning process.\n - Capability inventory: The skill is restricted to local data processing, specifically calculating view count medians and matching titles against regex patterns. No network, file-write, or shell execution capabilities are present in the provided scripts.\n
- Sanitization: Video titles are directly extracted and printed to the console for agent consumption without any sanitization, filtering, or escaping of potentially malicious prompt instructions.
Audit Metadata