crypto-sentiment
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from various external sources via web search results, which could contain malicious instructions designed to influence the agent's output.
- Ingestion points: The skill performs multiple
WebSearchqueries inSKILL.mdto collect data from Twitter, Reddit, news outlets, and GitHub repositories. - Boundary markers: The instructions lack clear delimiters or specific directives for the agent to ignore or isolate instructions potentially embedded within the retrieved search results.
- Capability inventory: The agent has the capability to write the aggregated data and its analysis into a markdown file (
CRYPTO-SENTIMENT-[TOKEN].md). - Sanitization: There is no mention of sanitizing, escaping, or validating the external content before it is processed and included in the final report.
Audit Metadata