stress-testing-game-concepts
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as natural-language game concepts, rule specifications, and telemetry data from prototypes as its primary function.
- Ingestion points: Ingests data from natural-language concepts and rules/spec documents (SKILL.md, Workflow Step 1).
- Boundary markers: There are no specific instructions to use delimiters or boundary markers to isolate user-supplied game rules from the system's analytical instructions.
- Capability inventory: The skill does not limit tool access in its frontmatter, meaning the agent's full suite of capabilities (e.g., file system interaction, shell access) could potentially be targeted if an injection successfully subverts the agent's logic.
- Sanitization: The skill instructions do not specify any validation, filtering, or escaping of the ingested game design content before processing it.
Audit Metadata