swarm-local-e2e
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands to manage the testing lifecycle, including process termination using
lsofandkill, removal of local database files withrm, and management of Docker containers. - [DATA_EXFILTRATION]: The skill accesses local environment files such as
.envand.env.docker-leadto retrieve configuration parameters likePORTandMCP_BASE_URL. These operations are limited to the local environment. - [DYNAMIC_EXECUTION]: The skill utilizes platform-specific execution features to run JavaScript snippets that interact with and verify responses from the local API server.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from a local API server, including agent statuses and task logs, which creates a potential surface for indirect prompt injection. 1. Ingestion points: Local API endpoints for agents, tasks, and session logs. 2. Boundary markers: None identified in the execution templates. 3. Capability inventory: Shell execution and JavaScript fetch capabilities. 4. Sanitization: No validation or sanitization is performed on the data retrieved from the local API server.
Audit Metadata