Backlog to Projects

Review active backlog items, identify natural groupings based on overlapping concerns, and propose projects with parallelism analysis. Present groupings to the user for approval, then create project folders for accepted groups.

When to Use

• Backlog has accumulated multiple items that need organizing
• User wants to plan next round of work from existing backlog
• After a batch of backlog items has been created
• Periodic backlog grooming to keep work organized

Workflow

Step 1: Inventory Active Backlog Items

Read all files in docs/backlog/ (excluding _archive/). For each item, extract:

• Title — from the H1 heading
• Area — which part of the codebase it touches (e.g. UI components, sidebar, file I/O, algorithm, stores)
• Type — the nature of the work: code change, UX exploration, investigation, refactor, bug fix
• Key files — from the References section
• Dependencies — explicit mentions of other backlog items or prerequisites
• Complexity signal — low (scoped code change), medium (design + implement), high (research + design + implement)

Present the inventory as a summary table:

| # | Item | Area | Type | Complexity |
|---|------|------|------|------------|
| 1 | ... | ... | ... | ... |

Step 2: Identify Natural Groupings

Cluster items using these signals (strongest to weakest):

1. Shared files — items that modify the same files or directories are strong candidates for the same project (avoids merge conflicts in parallel work)
2. Direct dependencies — if item A is a prerequisite for item B, they belong together
3. Conceptual cohesion — items addressing the same user-facing concern (e.g. "sidebar input polish" groups numeric stepper + toggle groups + selection)
4. Work type alignment — pure research items group separately from pure implementation items to allow different workflows

Items that don't cluster naturally become standalone projects.

Step 3: Assess Parallelism

For each proposed group, identify the files and directories it would touch. Two groups can run in parallel when their file footprints don't overlap. Flag conflicts explicitly:

• Parallel-safe — no shared files between groups
• Needs sequencing — shared files exist; note which files conflict and suggest ordering
• Can partially overlap — shared files are limited to read-only references or config; note the constraint

Present a parallelism matrix showing which groups can run concurrently.

Step 4: Present Proposal

Present the groupings with:

• Project name (kebab-case, 2-4 words)
• Included backlog items (by title)
• Rationale — why these items belong together
• Parallelism — which other groups it can run alongside
• Suggested approach — investigation-first vs. implementation-ready

Ask the user to approve, modify, or reject groupings before proceeding.

Step 5: Create Projects and Archive Backlog Items

For each approved grouping, create a project folder with a proposal. Follow the create-project conventions:

• Read docs/projects/TEMPLATES/PROPOSAL.template.md for the template
• Create docs/projects/<name>/proposal.md
• Fill in: title, date, status (Draft), overview, problem statement, proposed solution, scope, and related documents (linking to each backlog item)
• Check that the project name doesn't already exist in docs/projects/
• If a project already exists for a grouping, note it and skip creation

After project creation, archive the backlog items that were absorbed into projects:

• Move each item from docs/backlog/ to docs/backlog/_archive/
• Only archive items that belong to an approved and created project
• Items that remain ungrouped stay in the active backlog

The backlog is an inbox — once an idea graduates to a project proposal, the proposal becomes the source of truth. The archived backlog items remain accessible via the proposal's related documents links.

Step 6: Commit and Summarize

Stage and commit all new project folders and archived backlog items in a single commit. Present a final summary:

• Projects created (with paths)
• Projects that already existed (skipped)
• Parallelism overview — which can run concurrently
• Remaining ungrouped backlog items (if any)
• Suggested next steps per project (e.g. "needs investigation first" vs. "ready for dev plan")

Grouping Principles

Prefer fewer, cohesive projects over many tiny ones. A good project has 2-6 backlog items that share a clear theme. Single-item projects are fine for large or isolated work.

Don't force groupings. If items don't naturally cluster, leave them as standalone backlog items or single-item projects. Bad groupings create more coordination overhead than they save.

Consider the work type mix. A project mixing "quick code changes" with "needs UX research first" may stall the quick wins. Split if the research would block implementation items that are already clear.

Respect file boundaries for parallelism. The primary benefit of grouping is enabling parallel execution. If two items touch the same files, they should be in the same project to avoid merge conflicts.

Important Constraints

• Only read from docs/backlog/ (not docs/backlog/_archive/)
• Do not create plan.md or sessions/ — those come later
• Proposal content should reference backlog items as related documents using relative paths: ../../backlog/_archive/<filename>.md (since items are archived after project creation)
• Always get user approval before creating project folders
• Only archive backlog items after their project has been successfully created