The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): Hardcoded credentials admin:admin are used across multiple operational commands (backup, restore, status checks) for the Grafana service.
[COMMAND_EXECUTION] (HIGH): The skill utilizes high-privilege Docker commands including docker run --rm -v, docker volume rm, and docker compose. These allow for arbitrary file system modifications on the host through volume mounting and container lifecycle management.
[PROMPT_INJECTION] (HIGH): Category 8 (Indirect Prompt Injection) risk is high.
Ingestion points: Data is ingested via docker logs, Grafana search APIs (/api/search), and Loki log labels.
Boundary markers: None. Untrusted data from logs or dashboard titles is piped directly into other tools.
Capability inventory: High-privilege Docker operations and direct shell execution.
Sanitization: No sanitization or validation of data retrieved from the observability stack before it is processed or displayed to the agent.
[EXTERNAL_DOWNLOADS] (LOW): While the automated scan flagged a piped execution pattern, manual review confirms that the curl | python3 command in the 'Quick Commands' section correctly uses the -c flag for safe JSON processing rather than executing remote script content.

observability-control