The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill imports dashboard configurations from a remote Grafana API and converts them into Go code. This creates a surface where malicious content in the dashboard metadata could be processed by the agent or embedded in generated files.\n
Ingestion points: Dashboard data is fetched via gcx dev import dashboards from a live Grafana instance (SKILL.md).\n
Boundary markers: The skill does not describe any specific boundary markers or instructions to ignore embedded prompts in the imported data.\n
Capability inventory: The workflow involves writing Go files to the local filesystem and executing them using go run . (SKILL.md).\n
Sanitization: No explicit sanitization or validation of the dashboard JSON/YAML content is mentioned before conversion to Go code.\n- [DYNAMIC_EXECUTION]: The skill promotes a workflow where executable Go code is generated from external data and subsequently run.\n
Evidence: Instructions in the 'Workflow: Import → Edit → Push' section show the generation of code followed by execution via go run . (SKILL.md).

import-dashboards