The Agent Skills Directory

[SAFE]: The skill uses a secure authentication pattern by leveraging the --config option in curl. This ensures that sensitive API tokens are read directly from a configuration file rather than being passed as command-line arguments or environment variables, which prevents them from appearing in system process lists or logs.
[SAFE]: The skill includes specific jq logic to sanitize data before presentation. It redacts user authentication information (user:password) from datasource URLs, ensuring that credentials are not inadvertently exposed in the agent's output.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external API endpoints to retrieve model metrics, schemas, and task results, creating a surface for untrusted data ingestion.
Ingestion points: API response bodies from various $VM_ANOMALY_URL endpoints (SKILL.md).
Boundary markers: None explicitly defined for the API content interpolation, though responses are structured as JSON and parsed via jq.
Capability inventory: Operations are limited to network requests via curl and JSON processing via jq as specified in the allowed tools.
Sanitization: Proactive redaction of credentials in datasource URLs is implemented using jq regex substitution.

vmanomaly-query