The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to fetch a debugging guide from the official Azure GitHub repository (https://raw.githubusercontent.com/Azure/ARO-HCP/main/docs/ai/grafana-debugging.md) when needed. This documentation source belongs to a trusted organization.
[COMMAND_EXECUTION]: The skill executes local shell and PowerShell scripts that use the Azure CLI (az account get-access-token) to obtain authentication tokens and use curl or Invoke-WebRequest to communicate with Grafana API endpoints.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external Grafana instances, which constitutes a potential attack surface for indirect prompt injection.
Ingestion points: Results from Grafana API calls (datasources, metrics, and query results) are ingested into the agent context in scripts/list-datasources.sh, scripts/list-metrics.sh, and scripts/gquery.sh.
Boundary markers: No specific delimiters or warnings are used when outputting API results to the agent.
Capability inventory: The skill is capable of shell command execution and network requests.
Sanitization: The skill does not perform sanitization on the external API responses before they are presented to the agent.
[DATA_EXFILTRATION]: The skill scripts programmatically obtain an Azure access token and send it to the user-provided Grafana URL. While this is the intended functionality for accessing the Grafana API, it represents a vector where a token could be sent to an untrusted endpoint if the agent is provided with a malicious URL.

aro-grafana