cekura-create-agent
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to install dependencies (pip, npm), manage network tunnels (ngrok), and interact with the Cekura API via curl. These actions are performed to configure the user's environment for agent testing and observability as part of the intended setup flow.
- [EXTERNAL_DOWNLOADS]: Fetches and installs SDK packages (cekura[livekit], cekura[pipecat], @cekura/livekit) from official package registries. These are vendor-provided libraries required for integration. It also utilizes ngrok, a well-known service, for exposing local servers during the setup process.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied agent configurations, including system prompts and knowledge base files, which presents a potential surface for indirect prompt injection.
- Ingestion points: User-provided system prompts in phase4-description.md and knowledge base files in phase8-knowledge-base.md.
- Boundary markers: No explicit delimiters are used when handling ingested data.
- Capability inventory: The skill has the capability to install software, modify local source code files, and perform network requests to the vendor's API.
- Sanitization: User-supplied prompt content and document text are not sanitized before processing.
Audit Metadata