righter
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied UI copy strings during review or creation tasks. These strings are interpolated into agent prompts, which could contain instructions intended to influence the agent's behavior. \n
- Ingestion points: UI copy text provided by the user for review or generation (referenced in SKILL.md). \n
- Boundary markers: The skill uses 'Before:' and 'After:' blocks to separate inputs, which provides some context but does not fully mitigate adversarial input. \n
- Capability inventory: The agent is capable of executing local Node.js scripts and calling the vois_get_microcopy tool. \n
- Sanitization: No explicit sanitization or escaping of user input is defined before processing. \n- [COMMAND_EXECUTION]: Instructions in SKILL.md direct the agent to run local Node.js scripts (scripts/ari.mjs and scripts/check-slots.mjs) with user-provided text passed as command-line arguments. This pattern introduces a risk of command injection if the agent does not properly escape shell characters in the user-supplied strings.
Audit Metadata