trunk-quarantine

Investigate failing CI runs and quarantine flaky tests using Trunk's CI Autopilot MCP server.

518|49|Updated Apr 7, 2021
One-click install
npx skills add https://github.com/dxos/dxos --skill trunk-quarantine
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: trunk-quarantine
Source: https://github.com/dxos/dxos/tree/main/.agents/skills/trunk-quarantine
Command: npx skills add https://github.com/dxos/dxos --skill trunk-quarantine

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Flaky tests intermittently fail CI runs and block merges, forcing engineers to manually dig through raw CI logs to distinguish real failures from noise. This Skill uses Trunk's CI Autopilot MCP server to triage failures and quarantine statistically flaky tests.

Core Features & Use Cases

  • CI Failure Investigation: Pass a GitHub Actions workflow URL to investigate-ci-failure to pull Trunk's parsed test results, filter out already-quarantined flaky tests, and surface only real failures.
  • Flaky Test Diagnosis: Use search-test to find a test case ID by fuzzy name, then fix-flaky-test to fetch or trigger an investigation with evidence of flakiness.
  • Quarantine Workflow Guidance: Walks through the dashboard-only quarantine action (Quarantine → Always with a required comment), since no MCP tool performs the write.
  • Use Case: A PR check is red and you suspect flakiness. Give the workflow URL to the Skill, which identifies the failing test as known-flaky, retrieves its investigation evidence, and directs you to quarantine it in the Trunk dashboard so it stops blocking merges.

Quick Start

Investigate this failing GitHub Actions run and quarantine any flaky tests you find: <workflow URL>.

Frequently Asked Questions about trunk-quarantine

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

FAQPage Schema
How do I investigate a failing GitHub Actions CI run with Trunk?

Call the investigate-ci-failure tool with the workflow URL from gh run view or a PR check. It pulls Trunk's parsed test results, filters out already-quarantined flaky tests, and shows only real failures or build errors.

How do I quarantine a flaky test in Trunk?

Quarantining is a dashboard-only action with no MCP tool support. Find the test case ID with search-test, gather evidence via fix-flaky-test, then open the test's Trunk dashboard page and click Quarantine → Always with a required comment.

Does Trunk CI Autopilot require setup before use?

CI already uploads test results to Trunk, so no upload configuration is needed. Each user only needs a one-time per-user authentication with the trunk MCP server before its tools become available.

Why can't I confirm a test was quarantined through the MCP tools?

No MCP tool performs or verifies the quarantine write; it is gated by the repo's Manual Quarantine Permissions in the Trunk dashboard. Never report a test as quarantined until the user confirms the dashboard action completed.

What should I do if the CI job failed before tests ran?

If investigate-ci-failure reports a build or compile failure before tests executed, fall back to raw CI logs using gh run view <id> --log-failed to diagnose the underlying error.