consult-fable
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides templates for executing the
claudecommand-line interface to interact with a specific AI model. It assumes the tool is already installed in the user environment. - [SAFE_PRACTICE]: The provided CLI command templates explicitly include the
--disallowedTools Bash,Edit,Writeflag. This follows the principle of least privilege by ensuring the sub-process can only read information and cannot modify the system or execute further shell commands. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied text (the consultation question) and interpolates it into a shell command or subagent prompt.
- Ingestion points: User input provided as the
<question>argument inSKILL.md. - Boundary markers: The command uses the
--shell delimiter to separate options from positional arguments, which helps prevent argument injection. - Capability inventory: Execution of the
claudeCLI tool. - Sanitization: No specific sanitization of the user question is mentioned within the skill text.
Audit Metadata