linear-bug-triage

Deduplicate measured bug evidence and create or comment Linear triage issues.

34.0k|3.7k|Updated May 18, 2023
One-click install
npx skills add https://github.com/langfuse/langfuse --skill linear-bug-triage
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: linear-bug-triage
Source: https://github.com/langfuse/langfuse/tree/main/.agents/skills/linear-bug-triage
Command: npx skills add https://github.com/langfuse/langfuse --skill linear-bug-triage

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When monitoring surfaces a bug or regression with measured evidence, teams risk filing duplicate Linear issues, writing unsupported claims, or skipping human review. This Skill enforces a disciplined handoff: human approval first, then deduplication, then evidence-only Linear comments or new triage issues.

Core Features & Use Cases

  • Human Approval Gate: Presents a compact findings table (windows, deltas, Datadog links, proposed action) and stops unless the human explicitly approves the Linear handoff.
  • Deduplication Search: Searches open and recently closed Linear issues by error text, route, service, environment, and Datadog links before creating anything; adds an evidence comment to existing issues instead of duplicates.
  • Evidence-Only Issue Creation: Creates new issues in the Triage state with the bug label, a measured title, and a strict body template containing time windows, signals, Datadog links, and search terms—no root-cause guesses or fix suggestions.
  • Use Case: After a latency regression is confirmed in Datadog for an API route, use this Skill to check Linear for an existing issue, comment the new measurements on it, or file a clean triage issue with the measured delta.

Quick Start

Use the linear-bug-triage skill to deduplicate these measured regression findings and file or comment Linear triage issues after I approve the table.

Frequently Asked Questions about linear-bug-triage

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

FAQPage Schema
How do I create a Linear bug issue from measured monitoring evidence?

Gather recent and baseline time windows, measured deltas, affected services and routes, and Datadog links, then present them in a table for human approval. After approval, the Skill searches Linear for duplicates and creates a Triage-state issue with the bug label and an evidence-only body.

How does Linear issue deduplication work before filing a bug?

The Skill searches open Linear issues using exact error text, route, service, environment, monitor name, and Datadog link keywords, plus recently closed issues for recurring errors. If a match exists, it adds a concise evidence comment instead of creating a duplicate.

Can the Skill file a Linear issue without human approval?

No. It must first show a findings table with candidates, measurements, and proposed actions, and it stops unless the human explicitly approves. The gate is skipped only if a calling workflow already obtained explicit approval for the Linear handoff.

What evidence is required before commenting on a Linear issue?

Each candidate needs absolute recent and baseline windows with timezone, measured signals such as counts, rates, or latency percentiles, affected environments and routes, and Datadog links. Missing measurements must be stated as 'No measurements found'; guesses and unsupported claims are not allowed.

Does the Skill include root-cause analysis or fix suggestions in Linear issues?

No. New issues and comments are evidence-only: time windows, measured deltas, affected scope, Datadog links, and search terms. Fix suggestions, root-cause guesses, owner assignments, and next steps are excluded unless the user explicitly asks outside the Linear content.