cekura-metric-improvement
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a legitimate workflow for a specific product (Cekura). All mentioned domains (cekura.ai) and tool patterns belong to the author's verified infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill processes call transcripts and evaluation results which may contain untrusted data. However, the workflow is designed with human-in-the-loop verification, requiring the agent to ask the user for confirmation before submitting feedback or updating metrics, which effectively mitigates the risk of automated prompt injection.
- Ingestion points: Transcript data and metric results are retrieved from the Cekura observability API.
- Boundary markers: Instructions do not specify delimiters for external transcript content.
- Capability inventory: The skill can read call logs and modify metric configurations via the Cekura MCP tools.
- Sanitization: The skill relies on user-driven review and the 'Cost Guard' mechanism to prevent unintended bulk actions.
Audit Metadata