skills/bankrbot/skills/defi-native/Gen Agent Trust Hub

defi-native

Fail

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSPERSISTENCEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Automated security scanners have confirmed detections of malicious and phishing content within the documentation sources referenced in manifest.json. Flagged URLs include https://etherfi.gitbook.io/etherfi (Phishing), https://docs.elixir.xyz (Blacklist), and https://docs.eigencloud.xyz (Blacklist). The AI agent is explicitly instructed to visit and parse these locations. Furthermore, the manifest.json registry file itself has been flagged as malware (FileRepMalware) by reputation scanners.
  • [PERSISTENCE]: The skill's documentation in site/update/index.html provides instructions for setting up persistent shell command execution. It recommends using Claude Code 'SessionStart' hooks or system cron jobs to run git pull automatically for updates. This establishes a permanent execution path on the user's system that can be abused.
  • [REMOTE_CODE_EXECUTION]: Recommending automated git pull hooks for updates creates a remote code execution vector. If the upstream repository is compromised, malicious code could be silently downloaded and executed during future agent sessions.
  • [INDIRECT_PROMPT_INJECTION]: The skill architecture is based on fetching and processing untrusted content from over 120 external DeFi protocol documentation sites. This ingestion represents a large attack surface for indirect prompt injection, where an attacker could place malicious instructions on a monitored site to hijack the agent's behavior. While directives instruct the agent to treat this as data only, the risk of hijack remains.
Recommendations
  • CRITICAL: 1 file(s) identified as malware by FileRep - DO NOT USE
  • AI detected serious security threats
  • Contains 6 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 1, 2026, 10:17 PM