Devlab Cicd Onboard

DevLab CI/CD Onboard

Purpose

Use this Skill to introduce or evolve CI/CD in an application repository. It turns repository facts into a provider-neutral plan, then produces reviewed workflow and configuration changes for GitHub Actions, Jenkins, Woodpecker, or another established CI platform.

It is an engineering workflow, not a platform-administration runbook. It does not create accounts, change global CI settings, install plugins, or access credentials.

Inputs

• Repository path or source archive.
• Requested outcome: CI only, deployment pipeline, an environment change, or a migration between CI providers.
• Optional target environments and the CI provider already used by the team.

Outputs

• A short repository and delivery-path assessment.
• A reviewed pipeline design and project-local configuration template.
• Workflow files or a patch when the repository has enough confirmed context.
• A structured escalation when a CI platform, secret, runner, or deployment prerequisite needs an operator.

Workflow

1. Inspect the repository before selecting a provider: language, package manager, tests, build output, containerization, deployment target, and any existing CI files.
2. Select the smallest provider-compatible path. Preserve an existing working provider unless the request explicitly asks for migration. For a new public repository, GitHub Actions is the default recommendation.
3. Separate configuration by ownership:
   • versioned repository configuration contains only non-secret, portable values and placeholders;
   • CI secret stores hold tokens, SSH keys, registry credentials, and private endpoints;
   • platform-wide runners, plugins, credentials, and permissions remain operator-owned.
4. Produce a pipeline with explicit stages: dependency install, deterministic tests, build, artifact publication, optional deployment, and verification. A deployment must not happen until its target, credential reference, and rollback path are confirmed.
5. Offer quality gates as explicit, opt-in steps. Keep pull-request checks fast; prove each enabled gate with one intentional failure and one success.
6. Validate the generated YAML and scripts locally where possible. Then verify the provider run, artifact availability, and the deployed service contract separately.
7. Return an escalation instead of guessing when a required secret, runner capability, DNS rule, network path, approval, or deployment target is unknown.

Configuration Boundary

The Skill has no required global credentials. Store provider API settings in the provider's secret manager when an automation actually needs them. Keep project-specific, non-secret deployment settings in a reviewed file such as .platform/jenkins.yml or an equivalent provider-native configuration file.

Runtime dependencies

The bundled Jenkins-style validator runs in the target repository's Python virtual environment and requires PyYAML. The documented --schema path also requires jsonschema:

python -m pip install "PyYAML>=6.0" "jsonschema>=4.18"

The selected CI provider remains an external prerequisite. Its runner image, plugins, permissions, secret store, network access, and deployment tools are owned and configured by the repository or platform operator; they are not installed with this Skill.

Use the bundled scripts/validate-jenkins-yml.py only for the documented Jenkins-style project configuration. It validates structure; it does not prove that a remote host, credential, or CI service is reachable.

References

• Route table: provider and deployment selection.
• Escalation boundary: operator hand-off.
• Portable CI/CD patterns: provider-neutral templates and acceptance rules.
• Portable CI/CD patterns detail
• Configuration schema: optional Jenkins-style project configuration validation.

Constraints

• Never place credentials, private URLs, hostnames, or customer values in a tracked workflow, script, fixture, or example.
• Do not enable deployments, destructive cleanup, or production changes by default.
• Do not treat a green configuration check as deployment acceptance. Verify the actual artifact and service contract.
• Use a platform's native secret and approval mechanisms; do not invent a second credential store in the repository.
• Keep generated pipeline commands aligned with the repository's documented local install and build commands.

Human Decisions