The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or security vulnerabilities were detected. The skill implementation follows secure coding practices for handling environment variables and service credentials.
[COMMAND_EXECUTION]: The skill utilizes curl and jq within a Bash environment to perform necessary operational tasks. These commands are restricted to interacting with specific VictoriaMetrics API endpoints for data profiling, health checks, and configuration testing, all of which are directly relevant to the skill's primary purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external VictoriaMetrics API responses, creating a potential surface for indirect prompt injection. However, this is mitigated by the following factors: (1) Ingestion points: Metric characteristics and model schemas fetched from the configured $VM_ANOMALY_URL in SKILL.md; (2) Boundary markers: Not explicitly implemented for the API data itself, though the process is highly structured; (3) Capability inventory: Operations are limited to curl and jq for data processing; (4) Sanitization: The skill employs jq for generating structured JSON payloads and uses --data-urlencode for all query parameters, preventing basic command or injection confusion.

vmanomaly-config