The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches official Grafana tools and collectors, including the gcx CLI, the alloy collector, and the mcp-grafana server, from Grafana Labs' verified GitHub repositories and package registries.
[REMOTE_CODE_EXECUTION]: Instructions in PROVISIONING-AND-API.md describe installing the gcx CLI using a remote shell script (curl piped to sh). While this is a high-risk pattern, the source is an official Grafana Labs repository, which is a trusted organization.
[PRIVILEGE_ESCALATION]: The BEYLA.md and PYROSCOPE.md components require high-level system privileges (privileged: true and hostPID: true) to perform eBPF-based instrumentation on the host Linux kernel. Additionally, PROVISIONING-AND-API.md includes commands to bypass macOS Gatekeeper (xattr quarantine removal) for the gcx binary.
[COMMAND_EXECUTION]: The skill makes extensive use of CLI tools and APIs to perform administrative tasks such as resetting admin passwords, managing service accounts, and deploying stack configurations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes untrusted external data from logs, traces, and browser sessions.
Ingestion points: Browser session data ingested via gcx frontend sessions get in FARO-SESSIONS.md and log streams in LOKI.md.
Boundary markers: The skill provides an explicit warning to the user to treat session dumps as untrusted data that may contain malicious instructions or PII.
Capability inventory: The skill utilizes curl, kubectl, and the gcx CLI for stack management and resource provisioning.
Sanitization: Instructions emphasize manual review of evidence and the use of scoped selectors to minimize the ingestion of irrelevant or untrusted content.

grafana