CI

23 skills · 59 min

• Automatic Replication: Automatically monitors CI and immediately executes suggested test or lint commands locally upon failure.
• Real-time Monitoring: Aggregated status line for all concurrent workflows on the current branch.
• Fail-Fast Triage: Immediately stops on the first job failure to provide a structured report.

Workflow

1. CI Replicate (replicate) - DEFAULT

Use this as the primary path to monitor CI and automatically replicate failures locally for immediate triage and fixing.

• Behavior: When this workflow is triggered, the agent will monitor the CI and immediately and automatically execute all suggested test or lint commands (marked with 🚀) as soon as a failure is detected.
• Tool: node .gemini/skills/ci/scripts/ci.mjs [branch]
• Discovery: The script automatically finds the latest active or recent run for the branch. Do NOT manually search for run IDs.
• Goal: Reproduce the failure locally without manual intervention, then proceed to analyze and fix the code.

1. CI Status (status)

Use this when you have pushed changes and need to monitor the CI and reproduce any failures locally.

• Tool: node .gemini/skills/ci/scripts/ci.mjs [branch] [run_id]
• Discovery: The script automatically finds the latest active or recent run for the branch. You should NOT manually search for `run_id` using `gh run list` unless a specific historical run is requested. Simply provide the branch name.
• Step 1 (Monitor): Execute the tool with the branch name.
• Step 2 (Extract): Extract suggested `npm test` or `npm run lint` commands from the output (marked with 🚀).
• Step 3 (Reproduce): Execute those commands locally to confirm the failure.
• Behavior: It will poll every 15 seconds. If it detects a failure, it will exit with a structured report and provide the exact commands to run locally.

Failure Categories & Actions

• Test Failures: Agent should run the specific npm test -w <pkg> -- <path> command suggested.
• Lint Errors: Agent should run npm run lint:all or the specific package lint command.
• Build Errors: Agent should check tsc output or build logs to resolve compilation issues.
• Job Errors: Investigate gh run view --job <job_id> --log for infrastructure or setup failures.

Noise Filtering

The underlying scripts automatically filter noise (Git logs, NPM warnings, stack trace overhead). The agent should focus on the “Structured Failure Report” provided by the tool.