The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python build script (build.py), system utilities (rsync, git), and curl to interact with local observability services (Prometheus at 127.0.0.1:9090 and Loki at 127.0.0.1:3100) for metadata retrieval.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by ingesting untrusted data from external sources and processing it through its build pipeline.
Ingestion points: Reads dashboard specifications from observability/grafana/specs/ and external Grafana exports from /tmp/<uid>-export.json.
Boundary markers: The instructions do not define explicit delimiters or warnings to the agent to ignore instructions embedded within the processed JSON content.
Capability inventory: The skill can write files to the repository (dashboards/ and specs/), execute local Python scripts, and perform network requests to local services.
Sanitization: There is no evidence of sanitization or validation of the JSON content before it is processed or used to generate new dashboard files.
[INDIRECT_PROMPT_INJECTION]: The skill checks for an external configuration via the $POWERSUIT_OVERLAY environment variable, which instructs the agent to read and follow instructions from an external file path if present. This could allow external content to influence agent behavior.

dashboards-as-code