gamification-engine
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied descriptions of business behaviors and processes to generate structured code and data outputs. This creates a surface where malicious input could potentially influence the structure or content of generated SQL DDL, JSON schemas, or CSV files.
- Ingestion points: Step 1 and Step 2 capture user intent and current process descriptions (SKILL.md).
- Boundary markers: The instructions do not define specific delimiters (e.g., XML tags) or explicit instructions for the agent to ignore embedded commands within user-provided descriptions during artifact generation.
- Capability inventory: The skill generates SQL DDL statements, JSON schemas, and CSV data, and references external tools for Notion imports (SKILL.md Step 5).
- Sanitization: While the instructions advise 'material validation' and 'keeping prose outside the data', there are no explicit escaping or sanitization requirements for user-provided strings embedded in the final code templates.
- [DYNAMIC_EXECUTION]: The skill involves the dynamic generation of code-like artifacts (SQL and JSON schemas) based on variable user context. While these are generated as templates, the assembly of executable-adjacent content from untrusted input is a characteristic of dynamic execution environments.
- Evidence: Step 5 instructs the agent to 'derive the fields from the confirmed context' and provides templates for SQL, JSON, and CSV generation.
Audit Metadata