| name | lintel-board-prioritise |
| description | Groom and prioritise board work items — clean up done tasks, break down oversized work, and align priorities with the Lintel vision. |
/lintel-board-prioritise — Groom and Prioritise Board Work Items
Review all work items on the board, clean up completed/stale items, break down oversized tasks, and prioritise the backlog aligned with the Lintel vision.
Context: Lintel Vision
Lintel is the AI-human engineering platform — both measurement system and execution engine. Prioritisation should favour work that advances these pillars (in order):
- Event infrastructure — event store, bus, projections, subscriptions
- Collaboration & routing — teams, channels, integrations
- Guardrails & safety — cost limits, escalation, sandbox hardening
- Deployment & observability — deployments, incidents, feature flags
- Measurement & analytics — DORA metrics, agent accuracy, team velocity
- Continuous improvement — learning loops, automated remediation
Workflow Capacity Reference
A single feature_to_pr workflow run can handle work that fits in roughly one focused feature or bug fix — 11 stages: ingest → route → setup_workspace → research → plan → implement (up to 5 review cycles) → review → close. If a task requires multiple unrelated code changes across different packages, or touches more than ~3 files substantially, it should be broken down.
Steps
1. Fetch current state
Use MCP tools to gather data:
mcp__lintel__boards_list_boards — get all boards (needs a project_id; list projects first if needed)
mcp__lintel__work-items_list_work_items — get all work items
Present a summary table of current work items grouped by status.
2. Groom — remove completed work
Identify work items with terminal statuses (merged, closed) — these are done. Delete them with mcp__lintel__work-items_remove_work_item.
Also identify failed items — ask the user whether to retry (move back to open) or delete.
Report what was cleaned up.
3. Assess scope — break down oversized tasks
For each remaining open work item, evaluate whether it can be completed in a single feature_to_pr run:
Too big if:
- Description mentions multiple unrelated changes
- Touches more than 3 packages substantially
- Contains words like "and also", "plus", "as well as" suggesting bundled work
- Would require multiple PRs to review sensibly
For oversized items:
- Propose a breakdown into smaller work items (each achievable in one workflow run)
- Ask the user to confirm the breakdown
- On confirmation: create the new smaller work items with
mcp__lintel__work-items_create_work_item, then delete the original with mcp__lintel__work-items_remove_work_item
4. Prioritise
Rank all remaining open work items by:
- Vision alignment — how directly does this advance the Lintel vision pillars (see above)?
- Unblocking value — does this unblock other work items or capabilities?
- Foundation-first — lower-layer work (event infra, contracts) before higher-layer (UI, analytics)
- Risk reduction — security, data integrity, and correctness issues first
- User impact — features that make the platform usable for real workflows
Present the proposed priority order as a numbered list with brief rationale for each position.
5. Apply priority
After user confirms the priority order:
- Update
column_position on each work item using mcp__lintel__work-items_update_work_item to reflect the agreed priority (position 0 = highest priority)
6. Summary
Show the final board state: a clean table of prioritised work items with title, type, status, and position.
Important
- Always ask for user confirmation before deleting items or creating breakdowns
- Never delete
in_progress or in_review items — those have active workflows
- Use the Lintel MCP tools (
mcp__lintel__*), not HTTP requests
- If no project_id is known, use
mcp__lintel__projects_list_projects to find it first