shm-consumers
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents how to ingest data from an external local producer process, creating a potential surface for indirect prompt injection.
- Ingestion points: The wait_frame() method retrieves pixel data, and the on_message callback handles text messages sent from the producer process.
- Boundary markers: The provided snippets do not include boundary markers or instructions to treat incoming data as untrusted text.
- Capability inventory: The skill enables code that interacts with the local filesystem (shared memory regions) and performs inter-process communication via sockets or named pipes.
- Sanitization: While the documentation mentions protocol-level constraints (size limits and newline exclusion) for the reverse channel, it does not detail application-level sanitization of the received content.
Audit Metadata