commandkit-ai
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted AI-generated data into command handlers through
ctx.ai.params. It implements mitigation patterns such as Zod-based schema validation and explicit documentation warnings to treat all AI outputs as untrusted. - Ingestion points: Untrusted data enters the context via
ctx.ai.paramsas demonstrated inreferences/03-ai-command-schema.md. - Boundary markers: Uses Zod schemas to define expected data structures.
- Capability inventory: Handlers can perform Discord interactions (
ctx.message.reply) and execute custom tool logic. - Sanitization: Validation is handled by Zod schemas provided in the command configuration.
Audit Metadata