ghwf0-remote

Warn

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a shell script located at ~/.claude/scripts/ghwf/ghwf-daemon.sh using tmux. Since the contents of this script are not included in the skill, its exact behavior and security cannot be verified.
  • [COMMAND_EXECUTION]: The skill uses system commands like ps, grep, kill, and tmux to manage background processes and monitor the status of 'Claude' execution logs.
  • [PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface as it triggers actions based on external content from GitHub.
  • Ingestion points: The daemon monitors GitHub Issue and Pull Request labels, comments, and body descriptions for command triggers (e.g., ghwf:exec, ghwf:redo).
  • Boundary markers: There are no defined boundary markers or instructions to ignore malicious content within the issue body or comments being processed.
  • Capability inventory: The skill possesses the ability to start background sessions, execute local scripts, and terminate system processes.
  • Sanitization: While the documentation mentions a 'Collaborator-only' rule restricted to users with write access, the enforcement of this rule occurs within the external daemon script and cannot be validated by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 28, 2026, 11:34 PM