Files
magnus919_agent-skills/linear/SKILL.md
T
Magnus HedemarkandGitHub 672f6190b0 Add focused Linear Agent Skill and CLI
Adds a dependency-free task-oriented Linear GraphQL CLI, progressive-disclosure guidance, offline safety tests, public-schema validation, and SkillOpt-derived help and promotion-gate improvements.
2026-07-17 13:07:24 -04:00

5.9 KiB

name, description, license, compatibility, metadata, allowed-tools
name description license compatibility metadata allowed-tools
linear Manage Linear teams, projects, cycles, issues, comments, workflow state, and documents from a terminal through Linear's public GraphQL API. Use when a user asks to list, search, inspect, create, update, move, or comment on Linear work, or to find Linear documents. Do not use to embed a live agent inside Linear or to build an MCP integration. MIT Requires Python 3.8+; network access and a Linear personal API key or OAuth access token for live API requests.
service api graphql-docs oauth-docs rate-limit-docs agent-interaction-docs
Linear GraphQL https://linear.app/developers/graphql https://linear.app/developers/oauth-2-0-authentication https://linear.app/developers/rate-limiting https://linear.app/developers/agent-interaction
Bash Read

Linear

Use scripts/linear from this skill directory. It is a small, dependency-free wrapper around Linear's public GraphQL API, not an MCP server. Use command-specific --help rather than copying the full command reference into a response.

Setup

  1. Inspect the available command and relevant noun: scripts/linear --help and scripts/linear issue --help.
  2. For a live request, set exactly one credential in the process environment. Use LINEAR_API_KEY for a personal key or LINEAR_ACCESS_TOKEN for OAuth. Never print, persist, or place either in a command transcript.
  3. Begin with bounded discovery. Reads default to --limit 10; the maximum is 100.
  4. Output is always JSON: compact with --json, indented without it. --dry-run makes no network request and previews the operation.

Command Map

Need Command
Confirm current identity scripts/linear whoami --json
Discover teams scripts/linear team list --limit 20 --json
Narrow a known issue set scripts/linear issue list --team ENG --state "In Progress" --json
Find an issue by words scripts/linear issue search "customer import" --json
Read one known issue scripts/linear issue get ENG-42 --json
Read an issue with its project, cycle, hierarchy, comments, and relations scripts/linear issue get ENG-42 --detail --json
List or read projects scripts/linear project list --team ENG --json or scripts/linear project get "Roadmap" --json
List or read cycles scripts/linear cycle list --team ENG --json or scripts/linear cycle get UUID --json
Find documents by words scripts/linear document search roadmap --json
Read a document by UUID, slug, or URL scripts/linear document get REF --json
Use a documented unsupported GraphQL operation scripts/linear raw 'query { viewer { id } }' --json

Choose the Smallest Read

Situation Use
You know an issue identifier or UUID issue get
You have words but not an identifier issue search or document search
You need a bounded set with filters issue list, document list, project list, or cycle list
The task needs a documented operation outside this focused CLI raw with an explicit GraphQL query

raw is an escape hatch, not a replacement for normal commands. Keep its query narrow and use the official GraphQL documentation to confirm field names and permissions.

State Changes

Confirm the target, scope, and rollback path before acting. Read-only discovery may proceed without confirmation.

For issue create, issue update, issue move, issue comment, and raw GraphQL mutations:

  1. Identify the issue/team/state using a read command.
  2. State the exact intended change and recovery path to the user.
  3. Run the same command with --dry-run --json; this has no credentials or network requirement.
  4. After confirmation, rerun it with --confirm --json.
  5. Report the returned identifier and outcome without exposing credentials.

The --team filters for issue, project, and cycle lists require an exact team key. issue create resolves a team key or exact name before creation, resolves issue identifiers before comments or updates, and resolves a destination workflow state only within that issue's team. It does not guess IDs or workflow states. Load references/domain-and-workflows.md for safe mutation recipes and Linear workflow semantics.

Errors And Recovery

  • Missing credentials: export one supported environment variable only for the command session, or use --dry-run to inspect the request.
  • GraphQL error: the CLI writes Linear's first useful error message to stderr and exits nonzero, including when the HTTP status is 200. Check permissions, exact identifiers, and documented field availability.
  • Team ambiguity: use team list to choose an exact key/name; do not retry by guessing an ID.
  • State lookup failure: the error lists the available states for the issue's team. Use that exact name, or use documented raw GraphQL as the explicit escape hatch when the focused CLI cannot express the operation.
  • Limit failure: choose a value from 1 through 100. The CLI deliberately does not paginate automatically.
  • Rate limit or transport failure: wait and retry the same bounded read. Follow Linear's rate limiting guidance rather than adding a retry loop.

References

When you need... Load...
Linear's data model, workflow semantics, safe mutation recipes references/domain-and-workflows.md
GraphQL endpoint, auth, filters, pagination, errors, rate limits references/graphql-contract.md
CLI vs MCP vs raw GraphQL vs Agent Session decision references/integration-boundaries.md
Source URLs, access dates, schema verification procedure references/sources.md

Verification

Run the offline tests and repository validator after changes:

python3 -m unittest linear/tests/test_linear.py
ruby scripts/validate-skills.rb

When Not To Use

Use Linear's native MCP or agent-session/webhook API when the task is to embed a live agent inside Linear rather than operate Linear from a terminal.