Agent Skills: Linear Fix

Autonomously completes Linear tickets via git worktrees. Use when user says "fix ENG-123", "complete LIN-456", "work on linear ticket", provides a linear.app URL, or mentions a ticket ID pattern like ABC-1234. Creates git worktree, implements fix via autonomous planning/review loop, opens GitHub PR with full reasoning.

UncategorizedID: kvnxiao/.dotfiles/linear-fix

Install this agent skill to your local

pnpm dlx add-skill https://github.com/kvnxiao/.dotfiles/tree/HEAD/agent-configs/shared/skills/linear-fix

Skill Files

Browse the full folder contents for linear-fix.

Download Skill

Loading file tree…

agent-configs/shared/skills/linear-fix/SKILL.md

Skill Metadata

Name
linear-fix
Description
Autonomously completes Linear tickets via git worktrees. Use when user says "fix ENG-123", "complete LIN-456", "work on linear ticket", provides a linear.app URL, or mentions a ticket ID pattern like ABC-1234. Creates git worktree, implements fix via autonomous planning/review loop, opens GitHub PR with full reasoning.

Linear Fix

Autonomous Linear ticket completion: fetch issue → create worktree → plan → review → implement → PR.

CRITICAL: Create git worktree BEFORE implementation. Use ticket ID as worktree folder/branch name.

Usage

/linear-fix <ticket-id-or-url>
/linear-fix <ticket-id-or-url> --visual  # Enable visual testing

Workflow

Follow 6 steps (see ./references/workflow.md for full details):

  1. Gather Context - Extract ticket ID, fetch comprehensive details via linear-cli:
    • Always use --output json for all Linear CLI commands
    • Fetch issue details: linear-cli i get <id> --output json (title, description)
    • Fetch issue comments: linear-cli cm list <id> --output json
    • Check for uploaded files/attachments in both issue description AND comments
    • Use /linear-uploads skill to download any attachments (images, screenshots, files)
  2. Create Worktree - Use /git-worktree skill to create isolated worktree, cd into it, install deps
  3. Plan - Task tool with subagent_type=Plan, read ./references/planner-prompt.md for template
  4. Review - Task tool with subagent_type=feature-dev:code-architect, read ./references/reviewer-prompt.md for template
  5. Implement & Review - Execute approved plan, then subagent_type=feature-dev:code-reviewer, read ./references/code-reviewer-prompt.md. Visual validation (opt-in): runs subagent_type=squint:playwright-visual-validator in parallel if enabled via --visual flag, visual-test label, or [visual-test] marker in ticket
  6. Create PR - Commit, prepare PR body (fill template if exists), create PR. Read ./references/workflow.md Step 6 for PR body format

Autonomous Loops

Planning Loop (Steps 3-4):

  • Planner proposes 2-3 approaches → Architect reviews → revise if needed
  • Max 3 iterations; then select best available with caveats

Implementation Loop (Step 5):

  • Implement → Code reviewer reviews → fix if needed
  • Max 2 iterations; then proceed with caveats

IMPORTANT: Only proceed to Step 6 (Create PR) after code review is APPROVED.

PR Body Requirements (see workflow.md for details):

  1. Extract variables from Step 4 "Extract from approved review" section
  2. Always append the "🤖 Autonomous Planning Reasoning" collapsible section from Step 6b — use the EXACT template format