early-access-designer
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill manages the design of early access programs by ingesting external data, which introduces a surface for indirect prompt injection.
- Ingestion points: Data is pulled from manual exports or pasted records from launch platforms, web analytics tools (GA4), and mobile app store consoles.
- Boundary markers: The skill does not define structural delimiters for these inputs but provides a clear safety instruction.
- Capability inventory: The skill can write results to memory files and execute a local registry script to update event logs.
- Sanitization: Instructions explicitly mandate treating exports as untrusted and ignoring any instructions embedded within them.
- [COMMAND_EXECUTION]: The agent is instructed to execute a local script,
registry-events.py, to propose stage definitions to the system registry. - Evidence: The instructions direct the agent to submit definitions to
memory/events/launches.ndjsonvia an "authorizedoperation: proposerequest toregistry-events.py".
Audit Metadata