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 WebSearch queries in SKILL.md to 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
Risk Level
SAFE
Analyzed
Sep 3, 2026, 01:33 AM
Security Audit — agent-trust-hub — crypto-sentiment