cds-linear-issue-shaping

name	cds-linear-issue-shaping
description	Convert a plan of record into Linear-ready issues with dependencies, acceptance criteria, and sequencing. Use when a reviewed plan needs to become executable work in Linear.
disable-model-invocation	true

Linear Issue Shaping

Purpose

Convert a plan of record into well-shaped Linear issues with the right granularity, sequencing, and acceptance criteria. Issues should be independently workable and traceable back to the plan.

When To Use

• A plan of record exists and has been reviewed.
• The plan's phases or components need to become Linear issues.
• Issues need proper dependencies, milestones, and acceptance criteria.

When Not To Use

• No plan exists yet. Use planning first.
• The plan has not been reviewed. Run plan-review first.
• The work is a single obvious task. Just create the issue directly.

Core Principles

1. One issue, one workable unit. Each issue should be completable in isolation by a single agent or developer. If an issue requires context from another unfinished issue to start, it has a dependency that must be declared.
2. Acceptance criteria from the plan, not invented. Issue acceptance criteria come directly from the plan's phases and acceptance criteria. Do not add criteria the plan does not support.
3. Dependencies are explicit. If issue B cannot start until issue A is done, declare the blocked-by relation. Do not rely on issue ordering alone.
4. Traceable to the plan. Every issue ties back to the plan or authority it was shaped from. When that plan is a file in this repository with a concrete path, use the Plan Reference section (see Issue Template). When there is no repo-linked plan, state the source briefly in Problem instead—do not add an empty or placeholder Plan Reference section.
5. Right-sized, not over-decomposed. Do not split work into trivially small issues. Each issue should represent a meaningful unit of progress that can be validated.

Linear Hierarchy

Linear provides these organizational levels. Use what fits the work.

• Initiative — for large bodies of work spanning multiple projects or phases. Create when the effort is substantial enough to warrant top-level tracking (e.g., "Maestro Service" not "Fix bug X"). Name initiatives durably — implementation details go in the description, not the title.
• Project — the container for issues. One project per body of work.
• Milestone — groups issues within a project by phase or layer. Use when the plan has natural groupings.
• Issue — the workable unit. Flat issues with blocked-by relations, not sub-issues, unless a single issue needs further breakdown.
• Sub-issue — use sparingly, only when a single issue is too large and needs internal decomposition.

Workflow

1. Read the plan in full. Understand the phases, acceptance criteria, dependencies, and deferred work.
2. Decide organizational structure. Does this work warrant an initiative? How many milestones? Present the proposed structure for confirmation before creating anything.
3. Identify issue boundaries. Map plan phases or components to issues. Each issue should be a workable unit with clear acceptance criteria.
4. Define the issue sequence. Order issues by dependency. Issues with no upstream dependencies come first.
5. Shape each issue. Write the issue using the prescribed template. Use --description-file for all descriptions.
6. Create in Linear. Create initiative (if applicable), milestones, then issues. Add blocked-by relations after all issues exist.
7. Verify the full set. Confirm the issue set covers the plan completely — no gaps, no duplicates, no orphaned issues.

Issue Template

Each issue description follows this structure.

Plan Reference: Include the ## Plan Reference section only when there is a plan file in this repository with an actual path you can link (e.g. docs/foo-plan.md). If the plan lives only outside the repo (Linear doc, Notion, chat, no checked-in plan), omit the entire Plan Reference section—no section header, no bullets, no placeholders.

When a repo-linked plan exists, insert Plan Reference immediately after Problem, using this shape:

## Plan Reference

- Plan: `<path to plan file in this repository>`
- Phase/Component: `<which part of the plan this implements>`
- Spec Sections: `<relevant spec sections, if applicable>`
- Reference Implementation: `<path to reference code, if applicable>`

Full template (Problem through Architecture Notes; insert the Plan Reference block from above only when applicable):

## Problem

What this issue solves or implements. One paragraph.

## Scope

What is included in this issue.

## Non-Goals

What is explicitly excluded from this issue.

## Acceptance Criteria

- [ ] <criterion from the plan>
- [ ] <criterion from the plan>

## Verification

How to prove the work is correct.

## Architecture Notes

Relevant design context from the plan. Package, types, API surface.

Linear CLI Reference

Create initiative (large bodies of work only)

linear initiative create \
  --name "<initiative name>" \
  --description "<description with implementation details>" \
  --status planned

linear initiative add-project <initiative-id-or-name> <project-slug>

Create milestones

linear milestone create \
  --project "<project slug>" \
  --name "<milestone name>"

Create issues

cat > /tmp/issue-desc.md <<'EOF'
<issue description from template>
EOF

linear issue create \
  --title "<issue title>" \
  --description-file /tmp/issue-desc.md \
  --project "<project slug>" \
  --milestone "<milestone name>" \
  --priority <1-4> \
  --state "Todo" \
  --no-interactive

Add dependencies

After all issues are created:

linear issue relation add <issue-id> blocked-by <blocking-issue-id>

Constraints On This Skill

• This skill shapes and creates issues only. It must never generate implementation code.
• Issue acceptance criteria must come from the plan. Do not invent criteria.
• Use --description-file for all issue descriptions. Never pass markdown inline.
• Present the proposed structure (initiative, milestones, issue list) for confirmation before creating anything in Linear.
• Do not create labels. Labels are managed manually in the Linear UI.