| name | blueprint |
| description | Turn a one-line objective into a multi-session implementation plan with cold-start briefs. Use when asked to 'blueprint', 'multi-session plan', 'long-term plan', 'construction plan', 'break into sessions', 'plan across sessions', 'multi-step project plan', 'session plan', or when a feature is too large for a single Claude session. Produces a plan where each step can be executed independently by a fresh agent with no prior context. |
| argument-hint | <objective> |
| allowed-tools | ["Read","Glob","Grep","Bash","Agent","AskUserQuestion","WebSearch","WebFetch"] |
/blueprint โ Multi-Session Construction Plan
Announcement: Begin with: "I'm using the blueprint skill to create a multi-session plan."
Purpose
Large features can't be completed in a single Claude session. This skill decomposes a high-level objective into a sequenced plan where each step:
- Has a cold-start context brief โ everything a fresh agent needs to execute without reading prior session history
- Has explicit dependencies โ which steps must complete first
- Has verification criteria โ how to confirm the step is done
Context discipline while drafting
Context may be compacted mid-plan. If you sense the window tightening, write the plan to disk incrementally (one step at a time) rather than building the whole blueprint in memory. Produce every step through to the end โ do not stop early because of token concerns. If compaction fires mid-draft, resume from the last written step.
When to Use
- Feature spans multiple domains (backend + mobile + infra)
- Implementation will realistically take 3+ sessions
- Multiple developers/agents need to coordinate
- User says "this is a big one" or "break this down for me"
Pipeline
Phase 1: Research (understand the landscape)
- Read relevant PRDs, JTBDs, and existing task docs
- Explore the codebase areas that will be touched
- Identify existing patterns to follow
- List unknowns and risks
When the objective spans multiple codebase areas (e.g., backend + mobile + infra), spawn one Agent subagent per area in the same turn rather than reading sequentially. Each subagent returns its findings; you synthesize.
Phase 2: Design (create the dependency DAG)
- Break the objective into 3-8 implementation steps
- Map dependencies between steps (which must come first?)
- Identify parallelizable steps (can run simultaneously)
- Estimate relative complexity (S/M/L) per step
Phase 3: Draft (write step documents)
For each step, produce:
## Step [N]: [Title]
### Cold-Start Brief
> A fresh Claude session should read ONLY this section to begin work.
**Objective**: [What this step accomplishes]
**Codebase entry points**: [Specific files/directories to read first]
**Patterns to follow**: [Existing code patterns this step should mirror]
**Key decisions already made**: [Design choices from Phase 2 that constrain this step]
**What exists before this step**: [State of the codebase when this step starts]
### Dependencies
- Blocked by: [Step X, Step Y] or "None โ can start immediately"
- Blocks: [Step Z]
### Scope
- **In scope**: [Specific deliverables]
- **Out of scope**: [What NOT to do in this step]
### Acceptance Criteria
- [ ] [Criterion 1 โ testable/verifiable]
- [ ] [Criterion 2]
- [ ] All tests pass: `[specific test command]`
- [ ] Type check passes: `{{TYPECHECK_CMD}}`
### Estimated Complexity
[S / M / L] โ [Brief justification]
### Suggested Task Command
`/ct` with this step's scope to create a full technical decomposition.
Phase 4: Review (adversarial check)
First, list every concern you noticed while drafting โ missing migrations, implicit env setup, unclear step boundaries, weak acceptance criteria โ including low-confidence ones. Then classify each into: Completeness / Independence / Ordering / Gaps / Rollback. Carry the low-confidence items into the Risks table below with explicit "(low confidence)" tags rather than dropping them silently.
After listing, verify against the checklist:
- Completeness: Does executing all steps achieve the original objective?
- Independence: Can each step truly be executed by a fresh agent with only its cold-start brief?
- Ordering: Are dependencies correctly captured? No circular dependencies?
- Gaps: Are there implicit steps (migrations, config changes, env setup) that should be explicit?
- Rollback: If step N fails, does it break steps 1..N-1?
Output
Save the blueprint to: docs/superpowers/plans/YYYY-MM-DD-<feature-slug>.md
# Blueprint: [Feature Name]
**Objective**: [The original one-line objective]
**Created**: [Date]
**Steps**: [N]
**Estimated sessions**: [N]
**Parallelizable steps**: [List]
## Dependency Graph
Step 1 โโโ Step 2 โโโ Step 4
โ
โโโ Step 3 โโโ Step 5
## Steps
[Step documents from Phase 3]
---
## Risks & Mitigations
| Risk | Impact | Mitigation |
|------|--------|------------|
| [Risk 1] | [H/M/L] | [Strategy] |
## Open Questions
- [Question that needs user input before execution]
Integration
- Each step can be turned into a
/ct task document for detailed decomposition
- Steps can reference existing task docs in
tasks/ if they already exist
- The blueprint itself is a living document โ update it as steps complete
Constraints
- Cold-start briefs are self-contained โ a fresh agent reads only its step's brief and can begin work. If you find yourself about to reference "see Step 2", inline the information instead.
- Keep steps focused โ each step should be completable in a single session (1-3 hours of agent work)
- Plan to the level each step needs right now โ Steps 1โ2 (starting soon) get full cold-start briefs. Steps 3+ get a one-paragraph objective and can be refined later. Stop adding detail once the next executor has enough to start.
- Output is a plan document, not code edits. The deliverable is the blueprint file;
/si executes each step in a later session. If the user asks you to "also start on step 1", respond with the plan first and offer to hand off to /si afterward.