vmanomaly-review
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Uses
curlandjqto interact with VictoriaMetrics vmanomaly API endpoints for health status, model schema discovery, and configuration validation. Shell commands are used to process these API responses for reporting. - [PROMPT_INJECTION]: The skill ingests and interprets untrusted data from user-supplied YAML configurations and external API responses. This data is used to generate summaries and hypotheses, creating an indirect prompt injection risk.
- Ingestion points: Local configuration YAML files and response bodies from the VictoriaMetrics API endpoints in
SKILL.md. - Boundary markers: No explicit delimiters are used to wrap or isolate the untrusted data from the agent's instructions during processing.
- Capability inventory: The agent can execute shell commands via
curlandjqand has read access to the local filesystem. - Sanitization: The skill lacks explicit sanitization or validation of the ingested content before the agent evaluates it for model-data fit and detection quality.
Audit Metadata