ask-cc
Warn
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill constructs a bash command using a heredoc (
cat <<'EOF') to interpolate user-provided questions and file paths. This pattern is vulnerable to command injection; if a user provides input containing the delimiterEOF, they can terminate the heredoc and append arbitrary shell commands to be executed by the agent. - [PROMPT_INJECTION]: The instructions explicitly direct the agent to bypass filesystem sandbox restrictions in order to access OAuth credentials stored in the macOS Keychain. Instructing an agent to circumvent security controls or operate outside its designated sandbox is a high-risk pattern.
- [COMMAND_EXECUTION]: The skill performs shell execution of the
claudeCLI. While this is the intended functionality, the lack of sanitization for the user-supplied arguments interpolated into the command string creates a significant attack surface.
Audit Metadata