Linear

Linear Issue Management

Track your work, link PRs to issues, update issues as you progress.

Installation

npm install -g linearis

Setup

Check if authentication is configured:

# Check file first (preferred)
cat ~/.linear_api_token 2>/dev/null || echo "No token file"

# Or environment variable
echo $LINEAR_API_TOKEN

If neither exists, guide the user:

1. Create API key at https://linear.app/settings/account/security
2. Save it (file is simplest):

   echo "lin_api_..." > ~/.linear_api_token
   

   Or add to shell profile: export LINEAR_API_TOKEN="lin_api_..."

Necessary Context

Before you use this skill for anything, run linearis usage to get an overview of all available commands. Note that linearis output is always JSON.

On first use, also fetch user context to enable filtering by assignee and team:

curl -s -X POST https://api.linear.app/graphql \
  -H "Content-Type: application/json" \
  -H "Authorization: $LINEAR_API_TOKEN" \
  -d '{"query": "{ viewer { id name teamMemberships { nodes { team { key name } } } } }"}' | jq

This gives you:

• User ID - for --assignee filtering ("show MY issues")
• Teams - keys like ENG, DEV for --team filtering

Also fetch available labels:

linearis labels list | jq '.labels[].name'

Keep this context for the session.

Link PR to Issue

Always link PRs to Linear issues: this creates traceability and auto-updates issue status when PRs merge.

Linear's GitHub integration handles this automatically when the issue ID appears in your branch name or PR title. Setup guide

Start With the Issue ID in Your Branch

When you begin work on an issue in a branch, make sure to bake the issue ID into your branch name:

# Get the branch name from Linear (includes issue ID)
linearis issues read ENG-123 | jq -r '.branchName'
# → feature/eng-123-fix-login-timeout

# Create the branch
git checkout -b $(linearis issues read ENG-123 | jq -r '.branchName')

Include the Issue ID When Creating PRs

Reinforce the link by including the issue ID in your PR:

# Use magic "Closes" in the PR body
gh pr create --title "Fix login timeout" --body "Description here...

Closes ENG-123"

Fallback: Link Retroactively via Comment

Already created a PR without the issue ID? Link it manually:

linearis comments create ENG-123 --body "PR: $(gh pr view --json url -q .url)"

My Issues

The daily driver. Show issues assigned to the user, filtered by status.

# Active work (in progress + todo)
linearis issues search "" --assignee <user-id> --status "In Progress,Todo"

# Only in progress, scoped to the ENG team
linearis issues search "" --assignee <user-id> --status "In Progress" --team ENG

Present results clearly: issue ID, title, status, team.

Create Issue

Quick capture of new work. The --team flag is required.

# Simple (assigned to yourself)
linearis issues create "Fix login timeout" --team ENG -a <user-id> -d "Users report session expires"

# With labels and priority (1=urgent, 4=low)
linearis issues create "Fix login timeout" --team ENG \
  -d "Users report session expires too quickly" \
  --labels "Bug" --priority 1

# Multi-line description using bash $'...' syntax
linearis issues create "Refactor auth module" --team ENG \
  -d $'## Problem\n\nAuth code is tangled.\n\n## Proposal\n\nExtract to separate service.'

# Sub-issue (use --parent-ticket)
linearis issues create "Write auth tests" --team ENG --parent-ticket ENG-123

Before creating, search to check for duplicates, and broader context.

Track Progress

Keep issues in sync with your work. Read issues before writing: never lose information.

Check Current State

Before updating, understand what's there:

# Full issue with description and comments
linearis issues read ENG-123

Summarize key information: title, status, assignee, description, recent comments, parent/child issues.

Update Status

Move issues through workflow stages:

linearis issues update ENG-123 --status "In Progress"
linearis issues update ENG-123 --status "In Review"
linearis issues update ENG-123 --status "Done"

Update Description (Checkboxes, Sections)

To update issues, update progress, checkboxes, assumptions, you must use a sequence of read → modify → write to ensure you don't lose information or details.

Use these three commands in sequence:

# Get a fresh view of issue and comments
linearis issues read ENG-123

# Modify whatever after you read
linear issues update ENG-123 --description "My new description..."

You can also add updates via commenting:

# Progress note
linearis comments create ENG-123 --body "This issue doesn't make any sense, sorry"

Use comments when:

• Reporting incremental progress
• Adding investigation notes
• Linking external resources (PRs, docs, logs)

Use description updates when:

• Checking off task checkboxes
• Updating acceptance criteria
• Correcting original assumptions

Combine updates to issue description and adding comments as you see fit.

Search

Search issues by title and description:

# Search by keyword (returns array, not object with .issues)
linearis issues search "login timeout" --team ENG | jq -r '.[] | "\(.identifier) [\(.state.name)] \(.title)"'

Fallback if keyword search returns nothing: List all team issues, then filter locally:

# Step 1: Fetch all team issues to a temp file
linearis issues search "" --team ENG > /tmp/issues.json

# Step 2: List identifier, title, status
jq -r '.[] | "\(.identifier) [\(.state.name)] \(.title)"' /tmp/issues.json

# Step 3: Read the specific issue you need
linearis issues read ENG-123

Note: searching without --team searches ALL workspace teams.

List Projects

See active initiatives and their status.

linearis projects list

Projects can be used with --project when creating issues or filtering searches.

Rules

1. Always check LINEAR_API_TOKEN before running commands
2. Fetch workspace context on first Linear interaction (user ID, teams, labels)
3. Use --assignee with user ID to filter "my" issues
4. --team is required for creating issues
5. Link PRs to issues - offer this when gh pr create succeeds
6. Search before creating - avoid duplicate issues
7. Present JSON output clearly - format as readable summaries, not raw JSON dumps