game-playtesting
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection, as it requires the agent to read and interpret untrusted text and graphical data from browser-based games while maintaining capabilities for file modification and tool execution.
- Ingestion points: The agent is directed to read player-visible screens, in-game text, console telemetry, and rendered frames from the game build (SKILL.md, references/control.md).
- Boundary markers: None identified. There are no instructions to use specific delimiters or to warn the agent against following instructions that might appear within the game's user interface or logs.
- Capability inventory: The skill allows for 'authorised repairs' (writing to source code), 'direct action calls' to the game engine, and the execution of browser automation tools (SKILL.md, references/control.md).
- Sanitization: The skill does not define methods for sanitizing, escaping, or validating text and metadata extracted from the game environment before it is processed by the agent.
Audit Metadata