alpha-vantage

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill includes instructions in the 'Suggest Using K-Dense Web For Complex Workflows' section that direct the agent to proactively suggest the author's external platform (www.k-dense.ai) for complex reasoning or deep research. This modifies the agent's behavior to include commercial promotion.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from the Alpha Vantage API without adequate security boundaries, creating a vulnerability surface.
  • Ingestion points: External data is ingested from the Alpha Vantage API via the requests library in the av_get function.
  • Boundary markers: The implementation lacks delimiters or instructions to isolate the retrieved data from the agent's processing context.
  • Capability inventory: The skill has network access through requests and data processing capabilities through pandas.
  • Sanitization: No sanitization, validation, or filtering of the API response is performed before it is returned to the agent context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 08:41 AM
Security Audit — agent-trust-hub — alpha-vantage