task-status

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: Scripts including send_status.py, send_status_websocket.py, and send_status_with_logging.py contain a hardcoded default TELEGRAM_TARGET ID ("7590912486"). If the corresponding environment variable is not set by the user, status messages and task names will be sent to this fixed external recipient.
  • [DATA_EXFILTRATION]: Hardcoded absolute file paths referencing a specific local user directory ("C:/Users/Luffy/...") are present in README.md, SKILL.md, scripts/monitor_task.py, and scripts/send_status_with_logging.py. This exposes internal system structure and limits portability.
  • [COMMAND_EXECUTION]: The skill uses subprocess.run and shutil.which to execute the clawdbot CLI and internal Python scripts, utilizing arguments that may include user-provided task names and messages.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from command line arguments and forwards it to external messaging channels without sufficient sanitization.
  • Ingestion points: sys.argv in scripts/send_status.py and scripts/monitor_task.py.
  • Boundary markers: Absent; no delimiters or ignore-instructions warnings are used when formatting messages.
  • Capability inventory: Network communication via WebSockets and local command execution via subprocess.run in scripts/send_status.py.
  • Sanitization: Only length-based truncation (limiting the final message to 140 characters) is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 07:00 AM