onestore-web-sdk
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill incorporates the official ONE store H5 Game SDK by importing it from the vendor's CDN (h5sdk.onestore.net). This is a legitimate and expected behavior for a platform-specific integration tool.
- [INDIRECT_PROMPT_INJECTION]: The skill contains an injection surface where user-supplied configuration data (e.g., placement IDs, product IDs) is ingested via a form and interpolated into generated code. Evidence: 1. Ingestion points: Fill-in form in SKILL.md. 2. Boundary markers: Instructions provided in SKILL.md to guide the agent. 3. Capability inventory: Modification of game source files as defined in the core rules. 4. Sanitization: Not explicitly implemented for the configuration strings. Despite the surface, this is a standard configuration pattern for developer tools.
- [COMMAND_EXECUTION]: The skill suggests using diagnostic commands like curl -sI to inspect HTTP headers of the game's hosting server to verify iframe embedding requirements (evidence in references/diagnose.md). This is a benign use of system tools for debugging.
- [SAFE]: The skill promotes secure development practices by instructing the agent to include hooks for server-side verification of purchase tokens and ad rewards, rather than relying on client-side signals.
Audit Metadata