tavern-card-builder
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including novel source fragments, notes, and research materials, to generate roleplay card content. This creates a surface for indirect prompt injection where instructions embedded in source materials could attempt to influence the agent's behavior during the authoring process.
- Ingestion points: Materials are ingested via the material provenance and discovery workflows (documented in references/material-provenance.md and references/authoring-workflow.md).
- Boundary markers: The skill uses structured JSON blocks (e.g., twa-materials, twa-entries) and distinct Markdown headers to separate user-provided data from system instructions.
- Capability inventory: The skill interacts with other specialist skills (e.g., $sillytavern-card-pipeline, $sillytavern-api-reference) and includes a local Node.js validation script (scripts/validate-creative-authority.mjs).
- Sanitization: The skill employs rigorous validation checks (references/validation.md) and a dedicated script to verify the structural integrity of project files before processing.
Audit Metadata