compile-generation-prompts

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill operates as a pure data transformation engine with strictly defined boundaries. It explicitly prohibits the invention of dialogue or direct calls to external generation APIs, ensuring that it remains a tool for data compilation rather than autonomous action.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes external production artifacts (story_blueprint, production_assets, shot_plan) to generate prompts. This risk is assessed as safe due to the implementation of multiple validation layers and quality gates.
  • Ingestion points: Data entering the payload object via schemas/input.schema.json.
  • Boundary markers: The workflow instructions in SKILL.md and references/field-guide.md require freezing story facts and building a semantic representation, which acts as a logical delimiter.
  • Capability inventory: The skill has no access to sensitive files, network operations (beyond schema references), or shell execution capabilities.
  • Sanitization: Employs 'quality gate' checks as detailed in references/field-guide.md to block requests with missing assets or invalid parameters.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:36 AM
Security Audit — agent-trust-hub — compile-generation-prompts