Optimizes CircleCI configuration files for faster, more reliable CI pipelines.
npx skills add https://github.com/openai/plugins --skill circleci-config
Please help me install this Agent Skill. Skill: circleci-config Source: https://github.com/openai/plugins/tree/main/plugins/circleci/skills/config Command: npx skills add https://github.com/openai/plugins --skill circleci-config
CircleCI pipelines often suffer from slow runtimes, flaky behavior, and wasted compute caused by poor cache keys, oversized workspaces, missing test metadata, and redundant jobs. This Skill diagnoses those configuration bottlenecks and applies targeted fixes to .circleci/config.yml without changing product behavior.
save_cache/restore_cache steps.store_test_results, JUnit XML quality, circleci tests run, and timing-based parallelism for faster, more visible test suites.package-lock.json checksums, removes dead workspace persistence, and enables timing-based test splitting.Analyze my .circleci/config.yml and reduce the pipeline runtime by fixing caching, parallelism, and workflow filters.
High-intent search queries and answers about installing and using this skill.
Start by baselining the top longest and flakiest jobs, then remove duplicate jobs, tighten branch/tag filters, and fix cache keys to use lockfile checksums. Apply parallelism only to proven bottlenecks and validate impact against before metrics.
Replace volatile keys like {{ epoch }} or commit SHAs with versioned lockfile checksums such as v1-npm-{{ checksum "package-lock.json" }}. Cache the package manager's download store rather than broad directories like node_modules.
Use caches for reusable dependency data that speeds up future runs, workspaces for data handed off between jobs in the same workflow, and artifacts for outputs humans inspect later like reports or logs. Each mechanism has a distinct consumer.
First ensure store_test_results uploads valid JUnit XML with file or classname attributes, then use circleci tests run --split-by=timings to distribute tests. Only increase parallelism once timing data is reliable.
The restored path may not be read by any later command, or the cache may be broader than needed, making restore time exceed the savings. Audit each restore_cache step to confirm a downstream command actually consumes the restored data.
Avoid increasing parallelism when test volume is low or timing data is unreliable, since fan-out overhead can outweigh gains. Parallelize only measured bottlenecks and keep the workflow graph readable and deterministic.
kubiyabotManage CircleCI v2 pipelines, workflows, jobs, and project configurations via API.
soferMonitor deployed applications with health checks, metrics collection, and backlog issue generation.
roborev-devFetch a roborev review by job_id and implement its fixes.
ArabelaTsoMigrate build systems and CI/CD configurations between platforms like Maven to Gradle.