twitter-account-manager
Pass
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill contains a detailed list of prompt injection patterns (e.g., 'ignore all instructions', 'DAN mode'). Static analysis flagged these as suspicious, but technical review confirms they are part of a multi-layer defensive filtering system designed to protect the agent from hostile data encountered on social media.
- [EXTERNAL_DOWNLOADS]: The skill uses the Playwright library which downloads browser binaries (Chromium) during installation. This is standard behavior for the tool's core functionality of browser automation.
- [COMMAND_EXECUTION]: The skill instructions include commands for managing the automation daemon (start, stop, status) and installing system-level service units (launchd/systemd). These are implemented as user-level persistence mechanisms and require explicit user execution.
- [DATA_EXFILTRATION]: The skill transmits draft content and status flags to a user-configured Telegram bot for manual approval. This represents a legitimate human-in-the-loop channel rather than unauthorized exfiltration. Browser cookies are stored locally and used only for authenticating with the intended service (x.com).
Audit Metadata