twanahc/claude-linear-skill
5 stars · Last commit 2026-02-28
Claude Code skill for Linear — browse, triage, plan, propose improvements, and implement issues via subagent workflows. Multi-agent coordination via status signals, mandatory worktrees, and superpowers integration.
README preview
# Claude Code Linear Skill A Claude Code skill for working with Linear issues directly from your terminal — browse your backlog, triage, plan implementations, and execute, all through natural conversation. ## Why a Skill Instead of MCP? There are broadly two ways to give Claude Code access to external services: **MCP servers** and **skills**. This project takes the skill approach, and it's a deliberate choice. ### How MCP Works An MCP (Model Context Protocol) server registers tools globally. Once configured, every single turn of every conversation includes the tool definitions in the context — whether you're working on Linear issues or not. For a service like Linear with 8+ operations, that's a meaningful chunk of tokens consumed on every API call, even when you're just asking Claude to fix a typo. ### How This Skill Works The skill is only loaded when invoked (`/linear`). When it's not active, it costs zero tokens. The orchestrator skill spawns **subagents** for each phase (triage, planning, implementation), so the main conversation context stays clean even when processing many issues in a single session. ### Trade-offs | | MCP Server | Skill | |---|---|---|