The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided configuration files (mirrord.json) and extra_config inputs, which creates a potential surface for indirect prompt injection if the source content is untrusted.
Ingestion points: External configuration file mirrord.json and the extra_config input parameter used in the GitHub Action.
Boundary markers: The skill includes a dedicated Security Boundaries section that explicitly instructs the agent to treat these inputs as data rather than instructions.
Capability inventory: The skill utilizes the mirrord CLI to deploy images and manage traffic routing within Kubernetes clusters.
Sanitization: The instructions command the agent not to execute shell commands derived from these values and to review extra_config for unintended overrides.
[COMMAND_EXECUTION]: The skill provides numerous examples of using the mirrord preview CLI for cluster operations.
Includes commands for starting, stopping, and checking the status of preview pods (mirrord preview start/stop/status).
Provides Kubernetes resource definitions (RBAC) for managing identities and tokens for CI integration.
[EXTERNAL_DOWNLOADS]: The skill references several external resources and tools provided by the vendor.
GitHub Action: Uses metalbear-co/mirrord-preview for CI automation.
Helm Charts: References metalbear/mirrord-operator-share-ingress for deploying ingress components.
Documentation: Links to official documentation at metalbear.com and reference workflows in the metalbear-co/playground repository.

mirrord-prev-env