The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill handles external data and repository content, presenting a potential surface for injection. Ingestion points: External documentation fetched via sandbase_discover (references/sandbase-api-map.md). Boundary markers: Instructions to 'preserve the user's chosen stack' and 'authorization boundaries' (SKILL.md). Capability inventory: File modifications and external tool execution via sandbase_run (SKILL.md). Sanitization: Mandatory verification with tests and source comparison.
[SAFE]: Integration with SandBase tools is transparent and restricted to the vendor's own services. The workflow prioritizes user consent and verification, mitigating risks associated with automated configuration changes.

grafana-dashboards