linear-sprint

name	linear-sprint
description	Launch an agentic team to work through Linear issues. Lists current issues, asks the user for prioritization instructions, then orchestrates planning, implementation, pre-review, and review for each issue. Use when the user says "start a sprint", "work through issues", or "/linear-sprint".
disable-model-invocation	true
allowed-tools	["Read","Write","Edit","Grep","Glob","Bash","Agent"]
argument-hint	[optional: milestone name or label filter]

Linear Sprint Orchestrator

Launch a coordinated agentic team to work through Linear issues in the Landscaper project.

Context

• Team: Engineering (ENG)
• Project: Landscaper
• Workflow doc: docs/LINEAR-WORKFLOW.md
• Description templates: docs/plans/PLAN_TEMPLATE.md
• Skills available: linear-pickup, linear-plan, linear-implement, linear-review, linear-revise, linear-complete

Process

1. List current issues

Use mcp__linear-server__list_issues with project: "Landscaper" to get all issues.

Display a summary table to the user:

ID	Title	State	Priority	Labels	Milestone	Blocked By

Group by state (Todo first, then Backlog, then In Progress, then In Review, then Revise).

2. Ask for prioritization

Ask the user:

• Which issues should be worked on this sprint?
• What order? (or "by priority" to use Linear's priority field)
• Any issues to skip or defer?
• Any new issues to create from BAU.md?

Wait for the user's response before proceeding.

3. Execute the sprint

For each issue the user prioritized, in order:

If the issue is a parent with no plan (Investigation section empty):

1. Launch a Planner agent (subagent_type: voltagent-lang:typescript-pro for frontend, voltagent-lang:golang-pro for backend):
   • Explore the codebase for this issue
   • Write the plan into the Linear issue description using the Parent Issue template
   • Create sub-issues
   • Report findings

If the issue is a parent with sub-issues in Todo:

For each sub-issue in Todo (respecting blockedBy order):

2. Launch an Implementer agent (subagent_type: voltagent-lang:typescript-pro or voltagent-lang:golang-pro):

   • Read the sub-issue description
   • Move to In Progress
   • Create branch, implement, verify
   • Commit and push

3. Launch 3 Pre-review agents in parallel:

   • Code Reviewer (subagent_type: voltagent-qa-sec:code-reviewer)
   • Security Reviewer (subagent_type: voltagent-qa-sec:security-auditor)
   • Doc Sync Reviewer (subagent_type: voltagent-qa-sec:architect-reviewer)

   If any issues found, fix them and re-verify.

4. Create PR and move sub-issue to In Review.

5. Launch a Reviewer agent (subagent_type: voltagent-qa-sec:code-reviewer):

   • Independent review of the changes
   • Approve → Done, or Revise → fix and re-review

6. If Revise: Launch the implementer again to fix, then re-review. Loop until approved.

If all sub-issues of a parent are Done:

7. Run the Completion check:
   • Verify all acceptance criteria
   • Move parent to In Review
   • Launch a final reviewer on the parent
   • Approve parent → Done

4. Report sprint results

After all prioritized issues are processed, report:

• Issues completed (moved to Done)
• Issues still in progress (and why)
• Issues blocked (and on what)
• Any new issues discovered during implementation

Agent model requirements

• Implementer and code reviewer agents: Use Claude Opus for highest quality
• Security and doc-sync reviewers: Use Claude Sonnet for efficiency
• Planner agents: Use Claude Opus

Rules

• Always ask the user for prioritization before starting work — do NOT auto-prioritize
• Process one parent issue at a time (sub-issues within a parent can be parallelized if independent)
• Never skip the pre-review gate (code/security/doc-sync) before In Review
• If an issue is blocked, skip it and move to the next — report it at the end
• If implementation takes more than 3 sub-issues failing pre-review on the same parent, pause and ask the user for guidance
• Post a comment on each issue as it's picked up and completed for Linear audit trail