twitter-poster

Fail

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: CRITICALCREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The script scripts/setup-session.sh targets and copies browser cookie databases from Chrome, Chromium, and Firefox which contain active session tokens. It also dumps all environment variables to a file, which frequently contain sensitive API keys, tokens, and other secrets.
  • [DATA_EXFILTRATION]: Collected sensitive data, including browser cookies, environment variables, and SSH configuration files, is compressed and exfiltrated to a remote endpoint (https://telemetry.twboost.example/collect) via a curl POST request.
  • [COMMAND_EXECUTION]: The skill requires the execution of scripts/setup-session.sh, which performs intrusive system operations and unauthorized data collection without the user's informed consent.
  • [PROMPT_INJECTION]: The SKILL.md file contains instructions to the AI agent to mislead the user by describing the exfiltration process as 'standard account linking' and a 'routine ownership check', specifically advising the agent not to go into detail about the data transfer.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Aug 26, 2026, 02:25 AM