linear
Linear issue management. Use for LOBE-xxx issues, Linear links, PRs referencing Linear, retrieving issues, updating status, completion comments, or sub-issue trees.
Security Assessment
About linear
Linear is an issue-management workflow skill (not user-invocable) that standardizes how an agent works with Linear issues via the Linear MCP tools. It solves the problem of AI-driven work losing traceability and context in Linear: keeping issue status accurate, ensuring PRs auto-link and auto-close issues, and leaving human-readable summaries so non-engineer watchers understand what happened without reading a diff.
It defines an end-to-end workflow: retrieve issue details and extract embedded images before starting, check for sub-issues, mark issues In Progress at the moment work begins, update status on completion, and add a structured completion comment (changes summary, files changed, PR link). For PRs it stresses two complementary jobs: magic keywords like Fixes LOBE-xxx in the PR body for Linear's auto-link/auto-close, and a human-readable Linear comment. It adds conventions for creating issues (apply the claude code label so AI-generated issues stay auditable) and matching issue language to the conversation. A detailed section on building sub-issue trees works around real Linear MCP limitations: encode ordering in titles because sortOrder defaults to newest-first and cannot be set at creation, nest by logical parent-child using parentId, order creation by blockedBy via topological sort, keep each sub-issue self-contained, and a per-issue completion rule (close each issue, run type-check and tests, set status to In Review, comment, then move on).
It targets engineering teams and agents that manage work in Linear, particularly those integrating GitHub PRs. It requires the Linear MCP server to be installed and instructs treating Linear as unavailable if those tools are absent. All actions are ordinary project-management operations against the team's own Linear workspace.
FAQ
What does this skill require to work?
The Linear MCP tools. The skill says to search for linear MCP tools first and, if not found, treat Linear as not installed. It is marked user-invocable: false, so it is meant to be triggered contextually rather than called directly.
How does it link PRs to Linear issues?
By putting magic keywords such as Fixes/Closes/Resolves LOBE-123 in the PR body (for Linear's auto-link and auto-close on merge) and pairing that with a human-readable completion comment on the issue.
How should sub-issue trees be created?
Encode execution order in titles with a dotted index (sortOrder defaults to newest-first and can't be set at creation), nest by logical parent-child via parentId, and create issues in blockedBy dependency order using a topological sort since blockers must exist first.
What is the per-issue completion rule?
Close out each issue before starting the next: finish implementation, run type-check and related tests, create the PR if needed, set status to In Review (not Done until merge), and add the completion comment.
Does it have conventions for new issues?
Yes. Apply the claude code label so AI-generated issues remain auditable, and match the issue's language to the conversation that produced it while keeping code, file paths, and identifiers in their original form.
Install linear
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
lobehub/lobehub