The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by sampling live agent traffic and conversations to generate recommended production rules and guardrails.
Ingestion points: The skill uses gcx agento11y conversations search and gcx agento11y generations get to read live interaction data from the production stack.
Boundary markers: There are no explicit prompt delimiters mentioned for the sampled traffic data within the instructions.
Capability inventory: The skill utilizes Bash, Write, and Edit tools. It possesses the capability to create and update stack-level objects (rules, evaluators, and guards) via the gcx CLI.
Sanitization: The skill implements a mandatory manual review process. It drafts configurations to local YAML files and requires an explicit user confirmation for every individual write operation to the production stack.
[COMMAND_EXECUTION]: The skill uses the gcx (Grafana Cloud CLI) to perform discovery and apply configuration changes.
The skill instructions emphasize that all connections to the stack must go through the official gcx tool, which manages authentication securely.
The skill identifies and inventory existing rules and guards using gcx agento11y evaluators list, rules list, and guards list to avoid duplicate configurations.

agento11y-prod-setup