openclaw-ci-limits

Tune OpenClaw GitHub Actions runner-registration budgets and CI fanout caps.

388k|81.5k|Updated Nov 24, 2025
One-click install
npx skills add https://github.com/openclaw/openclaw --skill openclaw-ci-limits
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: openclaw-ci-limits
Source: https://github.com/openclaw/openclaw/tree/main/.agents/skills/openclaw-ci-limits
Command: npx skills add https://github.com/openclaw/openclaw --skill openclaw-ci-limits

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

OpenClaw CI runs against GitHub-hosted and Blacksmith runners with strict 5-minute registration buckets, and misjudging fanout causes queued jobs, throttled registrations, and slow merges. This Skill classifies the actual bottleneck before changing caps and enforces safe ramp-down/ramp-up rules.

Core Features & Use Cases

  • Bottleneck classification: Distinguishes runner-registration throttle, Blacksmith capacity, control-plane timeouts, unavailable labels, workflow dependency waits, and downstream ClawSweeper queue backlogs.
  • Registration budget math: Estimates worst-case Blacksmith registrations per 5-minute window from matrix rows, retries, and main parity slots, keeping planned load under about 60% of the live bucket.
  • Safe lever ordering: Preserves cancel-in-progress, canonical main parity slots, and CodeQL placement while recommending matrix reductions, runner right-sizing, and the OPENCLAW_CI_RUNNER_BACKEND circuit breaker.
  • Use Case: A maintainer notices PR runs queued for minutes. Use this Skill to inspect actions_runner_registration, classify the bottleneck, and propose a guarded change to strategy.max-parallel and runner labels.

Quick Start

Use the openclaw-ci-limits skill to diagnose why OpenClaw CI runs are queued and recommend a safe change to runner fanout or registration budget.

Frequently Asked Questions about openclaw-ci-limits

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce GitHub Actions runner registration throttling in OpenClaw CI?

Lower `strategy.max-parallel` for bursty Blacksmith matrices, preserve cancel-in-progress for superseded PR heads, and keep planned burst load under about 60% of the live `actions_runner_registration` bucket. Verify the bucket live before each tuning pass.

What is the difference between Blacksmith capacity issues and runner registration throttling?

Runner-registration throttling shows 403/429 or spam-style 422 responses from `generate-jitconfig` while core API quota stays healthy. Blacksmith capacity issues appear as concurrency caps or unavailable capacity on the Blacksmith dashboard and cannot be fixed by workflow fanout alone.

How do I switch OpenClaw CI to GitHub-hosted runners during a Blacksmith outage?

Set the repository variable `OPENCLAW_CI_RUNNER_BACKEND` to `github` using `gh variable set`. This routes configurable `ci.yml` jobs to hosted fallback labels and disables Blacksmith-only Docker and sticky steps. Delete the variable to restore Blacksmith-first routing.

Why are OpenClaw CI jobs queued even when runners appear available?

Run-level `queued` status hides whether the job waits on workflow dependencies or has no eligible runner. Compare `created_at`, `started_at`, `labels`, and `runner_name` per job to classify the wait before changing caps.

When should I not use the openclaw-ci-limits skill?

Do not use this skill for ordinary test failure triage, slow individual test runtime, or jobs that fail after starting. Use `$github:gh-fix-ci` or `$openclaw-testing` for real failing CI, and `$openclaw-test-performance` for test runtime tuning.