CI/CD AI Agent Directives

CI/CD is the fastest-moving part of an infrastructure estate, and failures there stop delivery faster than any outage in the cloud. These CI/CD directives give an AI agent focused tasks on the pipeline itself: diagnose a failed GitHub Actions run from the actual failing step, optimize slow workflows with caching and parallelism, scan for committed secrets, resolve GitOps sync conflicts from the diff, plan a safe multi-environment promotion with gates, generate pipeline documentation from the YAML, and tune Docker builds. Each directive is written against the real tools, so a cursor rules github actions or windsurf ci cd rules prompt runs with gh, actions YAML, and git rather than a generic guess. The output is consistently actionable: a root cause, a scoped fix, or an ordered plan, never a wall of logs. If you are collecting cursor rules ci cd, github actions directives, or ai agent devops directives for release engineering, this category is the set that shortens the gap between a red pipeline and a green one. Keep them as a repo-level rules file so the next failed build gets a reproducible diagnosis, not another manual log tail. They turn whatever goes wrong in delivery into a reproducible diagnosis instead of another manual tail of a failed build, so the team's time is spent fixing the pipeline rather than reading it.