grafana-loki
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted log data.
- Ingestion points: Log streams ingested through Loki push endpoints or collectors like Alloy as shown in SKILL.md.
- Boundary markers: LogQL provides structural separation for queries, but the skill does not define specific agent-level prompt boundaries for log content consumption.
- Capability inventory: Restricted to log aggregation, querying, and metadata management; no subprocess execution, file-system writing, or dangerous network capabilities are provided.
- Sanitization: The ASSESSMENT.md file explicitly includes redacting sensitive data before storage as a key review criterion.
Audit Metadata