rstack-eco-ci-debug

Diagnose Rstack ecosystem CI failures and attribute the true source PR or downstream change.

12.9k|843|Updated Apr 1, 2022
One-click install
npx skills add https://github.com/web-infra-dev/rspack --skill rstack-eco-ci-debug
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: rstack-eco-ci-debug
Source: https://github.com/web-infra-dev/rspack/tree/main/.agents/skills/rstack-eco-ci-debug
Command: npx skills add https://github.com/web-infra-dev/rspack --skill rstack-eco-ci-debug

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Rstack ecosystem CI suites can turn red for misleading reasons: the visible Rspack pivot commit is often not the real cause, since downstream dependency bumps, snapshot updates, flaky selectors, and release windows create false attributions. This Skill provides a disciplined two-phase workflow to find the actual source PR or downstream change instead of over-blaming the first red Rspack commit.

Core Features & Use Cases

  • Two-phase attribution workflow: Phase 1 locates the actual source PR, date window, or downstream change using fast-exit checks, green-to-red pivot analysis, and failure signature comparison; Phase 2 explains the technical mechanism behind a confirmed Rspack PR.
  • Specialized investigation tools: Includes a status helper script for parsing eco-ci status data, a canary date bisect procedure using @rspack-canary/core pnpm overrides, Rsbuild config debugging via DEBUG=rsbuild, daily automation triage, and guarded PR report commenting.
  • Use Case: A daily eco-ci run shows the modernjs suite newly failing at a Rspack commit. Use this Skill to check whether the same commit passed before, compare failure signatures against prior flaky history, bisect canary versions if the window is coarse, and produce an evidence-based triage report with a confident attribution.

Quick Start

Ask the agent to investigate why a specific Rspack eco-ci suite turned red and identify which PR or downstream change actually caused it, providing your local Rspack checkout path.

Frequently Asked Questions about rstack-eco-ci-debug

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

FAQPage Schema
How do I find which Rspack PR broke an eco-ci suite?

Identify the green-to-red pivot in the eco-ci status data, pull the failure logs for the current and pivot runs, and compare failure signatures. Apply fast-exit checks first: if the same Rspack commit appears in both green and red runs, the cause is downstream, not that commit.

How to bisect Rspack canary versions to locate a regression?

Use pnpm.overrides in the downstream project to pin @rspack/core to specific @rspack-canary/core versions, then binary-search by publish time between a known-good and known-bad canary. Verify each override with pnpm why before testing, and restore the downstream tree afterward.

Why does eco-ci blame a Rspack commit that did not cause the failure?

Eco-ci tests a combination of current downstream state plus the tested Rspack artifact, so downstream dependency bumps, snapshot changes, or flaky timeouts can create misleading pivots. The visible pivot commit is only surface attribution until failure signatures and diffs are compared.

When should I use DEBUG=rsbuild during eco-ci debugging?

Use it only when the attribution hypothesis depends on a config-gated option, plugin, loader, target, or mode being active in the failing case. Inspect the generated configs under dist/.rsbuild/ to confirm the relevant option is enabled before blaming a config-gated PR.

Can the skill post comments on the PR that caused a failure?

Yes, but only after strict attribution confirms a merged source PR, flaky and config-gated alternatives are ruled out, and the user explicitly approves posting. The comment includes a required agent marker, evidence links, and the checks performed.

What are the limitations of eco-ci failure attribution?

Attribution fails when evidence is insufficient, signatures conflict, or the same failure predates the candidate PR. In those cases the correct verdict is flaky, pre-existing, or inconclusive rather than forcing a weak PR attribution.