An opinionated, safety-focused Agent Skill for designing, auditing, implementing, and troubleshooting production delivery pipelines on Yandex Cloud.
The skill combines general continuous-delivery practices with Yandex-specific guidance for workload identity federation, Terraform, Container Registry, Compute, Instance Groups, Managed Kubernetes, Serverless, Application Load Balancer, Monitoring, Logging, and Audit Trails.
Install the repository with an Agent Skills-compatible client, or copy skills/yandex-cloud-cicd into one of the skill directories supported by your agent.
Examples:
npx skills add DanilTezin/yandex-cloud-cicd-skillgh skill install https://github.com/DanilTezin/yandex-cloud-cicd-skillExact installation support depends on the agent. The skill itself follows the open Agent Skills specification.
- audits existing GitHub Actions, deployment scripts, and Terraform layouts;
- designs staged application and infrastructure delivery pipelines;
- prefers short-lived Yandex Cloud credentials through workload identity federation;
- separates build, promotion, migration, deployment, verification, and recovery;
- selects rollout strategies for VMs, Instance Groups, Kubernetes, Serverless, and static assets;
- identifies common pipeline races, credential risks, mutable artifacts, and unsafe retries;
- uses official Yandex Cloud documentation and MCP servers as current references.
The skill does not silently deploy infrastructure. Read-only discovery is the default; external mutations retain the authorization boundary of the agent using it.
Run the deterministic checks:
python3 -m unittest discover -s tests -v
python3 scripts/validate_skill.pyThe bundled repository auditor can be run against another checkout:
python3 skills/yandex-cloud-cicd/scripts/audit_repository.py /path/to/repositoryThe first release covers GitHub Actions and the common Yandex Cloud deployment targets. Contributions for GitLab CI, policy-as-code integrations, and additional tested fixtures are welcome.
MIT