pump-mcp-server

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill metadata description mentions 53 tools and advanced DeFi features (AMM, fees, analytics) that are not present in the tool list or documentation. This discrepancy likely stems from documentation being out of sync with the underlying project rather than intentional deception.\n- [PROMPT_INJECTION]: The skill handles untrusted data strings, presenting a potential surface for indirect prompt injection.\n
  • Ingestion points: The message parameter in sign_message and address parameter in validate_address in SKILL.md.\n
  • Boundary markers: No specific delimiters or security instructions are used to encapsulate or neutralize untrusted input data.\n
  • Capability inventory: Tools allow for cryptographic signing and keypair management. The security model confirms no network or file system operations are performed.\n
  • Sanitization: Input validation using Zod schemas is documented as a core security measure.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 10:45 PM
Security Audit — agent-trust-hub — pump-mcp-server