Improve skill description for deploy.#173
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request updates the google-cicd-deploy skill definition by expanding its description and adding detailed trigger scenarios for GCS, GKE, and Cloud Run. It also introduces a critical requirement for skill activation prior to executing deployment-related tasks to ensure security protocols like secret scanning are followed. Review feedback suggests expanding the list of supported backend languages for Cloud Run and adding kubectl to the list of commands requiring prior activation to better reflect the skill's scope.
| - Any user request with the intent to get an application "up and running", "working", "hosted", or "online" on Google Cloud / GCP. | ||
| - Deploying static websites, uploading build artifacts/dist folders, or configuring hosting on Google Cloud Storage (GCS). | ||
| - Running containerized applications or Docker images on Google Kubernetes Engine (GKE), including creating/updating Kubernetes manifests (Deployment, Service). | ||
| - Deploying backend apps (Node.js, Python, Go) to Cloud Run, whether deploying from source code, container images, or using Cloud Native Buildpacks. |
There was a problem hiding this comment.
The list of supported languages for Cloud Run should be more inclusive to match the capabilities defined in the prompt (Workflow B), which supports Go, Node.js, Python, Java, .NET, Ruby, and PHP. Using "such as" or expanding the list helps ensure the skill is triggered for all supported language runtimes.
| - Deploying backend apps (Node.js, Python, Go) to Cloud Run, whether deploying from source code, container images, or using Cloud Native Buildpacks. | |
| - Deploying backend apps (such as Node.js, Python, Go, Java, .NET, Ruby, or PHP) to Cloud Run, whether deploying from source code, container images, or using Cloud Native Buildpacks. |
| - Deploying backend apps (Node.js, Python, Go) to Cloud Run, whether deploying from source code, container images, or using Cloud Native Buildpacks. | ||
| - Planning, troubleshooting, or analyzing the best Google Cloud deployment target for an application (even if the request has missing files or false assumptions). | ||
|
|
||
| **CRITICAL**: Activation is MANDATORY before writing any Dockerfiles, Kubernetes manifests, or running deployment shell commands (such as `gcloud storage`, `gsutil`, `gcloud run deploy`, or `gcloud deploy`). This guarantees compliance with required security protocols (like secret scanning) and prevents security leaks. |
There was a problem hiding this comment.
The list of deployment commands in the CRITICAL section should include kubectl, as it is the primary tool used for GKE deployments (Workflow D). Including it ensures the model recognizes that kubectl operations also require prior skill activation to trigger mandatory security protocols like secret scanning.
| **CRITICAL**: Activation is MANDATORY before writing any Dockerfiles, Kubernetes manifests, or running deployment shell commands (such as `gcloud storage`, `gsutil`, `gcloud run deploy`, or `gcloud deploy`). This guarantees compliance with required security protocols (like secret scanning) and prevents security leaks. | |
| **CRITICAL**: Activation is MANDATORY before writing any Dockerfiles, Kubernetes manifests, or running deployment shell commands (such as `gcloud storage`, `gsutil`, `gcloud run deploy`, `kubectl`, or `gcloud deploy`). This guarantees compliance with required security protocols (like secret scanning) and prevents security leaks. |
1 participant
It turns out that organizing the trigger scenarios in a bullet-list format can improve the skill activation.
Total pass rate for skill activation test raised from 78% to 96% for google-cicd-deploy.