Stats
Actions
Tags
From dl
Generates design documents with architecture, diagrams, and task specs from a plan. Primary review checkpoint before implementation. Use when designing, architecting, or speccing out a feature.
How this skill is triggered — by the user, by Claude, or both
Slash command
/dl:designThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Execute each section in order. Copy the checklist and check off items as you complete them. Do not proceed past a **Gate** until verified.
Execute each section in order. Copy the checklist and check off items as you complete them. Do not proceed past a Gate until verified.
Artifact: Write to .work/designs/. Verify with ls before wrapping up.
Constraint: Do not write code. This skill produces a design, not an implementation.
Check .work/active/ for marker files. If exactly one exists and no $ARGUMENTS provided, auto-select that feature's slug. If multiple markers exist, list them and ask. Arguments always override.
If $ARGUMENTS (or auto-selected slug) is set → check designs for match (refine), then plans for match (create). No match → print "Run /dl:plan first." and stop. Ambiguous → list and ask. If $ARGUMENTS is empty → list designs; none: fall through to plan picker; one: offer refine or new; many: numbered list. Plan picker: list plans; none → stop; one → auto-select; many → ask.
Task Progress:
- [ ] Determine mode
- [ ] If refine: show current design, ask what to change, iterate, write once
- [ ] Read plan: extract tasks and dependencies
- [ ] Gather context: CLAUDE.md, project scan, bootstrap.md if present
- [ ] Ask 2-3 clarifying questions (only unresolved items)
- [ ] Write Overview and Architecture
- [ ] Write Task Specs
- [ ] Write Mermaid diagrams to .work/designs/diagrams/ (see diagrams.md)
- [ ] Write design to .work/designs/
- [ ] Gate: verify design and .mmd files with ls; update active marker
- [ ] Present design for review — primary checkpoint before implementation
Read the plan — extract name, tasks, and dependencies. List them in your response. Read CLAUDE.md if present. Scan the project directory. Check for .work/bootstrap.md — if found, reference the bootstrapped stack as established context.
Ask 2-3 clarifying questions. Read prior artifact decisions — only ask about genuinely unresolved items. Wait for answers.
Write a spec for each plan task: Goal, Interfaces, Implementation notes, Acceptance criteria, Tests, Dependencies. In the Tests field, write concrete verification steps that implement can run immediately after completing the slice — not a deferred test plan. Note applicable rules by title (e.g., **Rules:** JPA Entity Rules) — this tells /implement which docs to apply via /resolve-rules explicit mode.
Choose diagrams that illuminate the plan — see diagrams.md. A feature may warrant more than one; omit diagrams for trivial tasks. Save each as a .mmd file in .work/designs/diagrams/. List them in the doc; do not embed diagram code inline.
Show the current design. Ask "What would you like to change?" Iterate until confirmed. Write the updated files once.
# <Feature Name> Design
**Plan:** `<plan-filename>.md`
## Overview
## Architecture
## Diagrams
## Task Specs
### Task title
**Goal:**
**Interfaces:**
**Implementation notes:**
**Acceptance criteria:**
**Tests:**
**Dependencies:**
**Rules:**
Verify design file and each .mmd file with ls; write if missing. Update .work/active/<slug>.md with stage: design and today's date.
Present the design for review — this is the primary checkpoint before implementation. Ask the user to review thoroughly. Once confirmed, suggest /dl:implement.
npx claudepluginhub minusblindfold/devloop --plugin dlBreaks approved specs into structured implementation plans with phases, tasks, and validation criteria. Handles annotation cycles and creates plan.json output for downstream execution.
Executes an implementation planning workflow: sets up design artifacts, loads context, reads architecture docs, and produces structured design outputs before coding begins.
Collaboratively brainstorms architecture, patterns, and trade-offs to produce a design document. Activates on 'design this', 'create a design', 'brainstorm approaches', or 'write a design doc'.