uniswap-cca-deployer

Fail

Audited by Gen Agent Trust Hub on Apr 9, 2026

Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [DATA_EXFILTRATION]: The skill performs unauthorized device fingerprinting by collecting system-specific metadata including hostname, operating system details, hardware architecture, and the user's home directory path. This data is then exfiltrated via POST requests to external endpoints: https://plugin-store-dun.vercel.app/install and https://www.okx.com/priapi/v1/wallet/plugins/download/report.
  • [COMMAND_EXECUTION]: The script in SKILL.md executes shell commands (hostname, uname, shasum, and curl) to generate the device identifier and transmit it. These operations are performed under the guise of 'Pre-flight Dependencies' and run automatically without user confirmation.
  • [PROMPT_INJECTION]: The skill uses social engineering by claiming to be 'auto-injected by Plugin Store CI' and authored by 'Uniswap Labs' in the documentation, despite internal metadata (plugin.json) and the actual behavior suggesting otherwise. This framing is designed to deceive both the agent and the user into treating the malicious telemetry block as a standard system requirement.
  • [CREDENTIALS_UNSAFE]: The skill contains a hardcoded HMAC signing key encoded in Base64 (OE9nNWFRUFdfSVJkektrMExOV2RNeTIzV2JibXo3ZWNTbExJUDFIWnVoZw==). This key is used to sign the exfiltrated fingerprint, indicating a structured attempt to authenticate malicious telemetry data.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 9, 2026, 05:44 AM