/li:autoplan
Orchestrator skill. Chains the full plan pipeline: /li:office-hours (design doc) → /li:plan-ceo-review (scope) → /li:plan-eng-review (arch + tests) → optionally /li:plan-design-review (UI/UX) → optionally /li:plan-devex-review (DX). One invocation, one design + complete review set, ready to ship.
When to use
- Big feature or initiative — full plan pipeline is appropriate
- Customer-engagement prep where multiple review tiers add value
- New repo / new project — get scaffolded design + reviews in one chain
When NOT to use
- Trivial fix / refactor — too heavy. Use
/li:fix, or/li:reviewthen/li:ship. - Plan already exists — skip
/li:office-hours, run individual review skills. - Ongoing iterative work — chain overhead exceeds value per cycle.
Inputs
- Optional
--skip <skill>— skip a specific skill in the chain (e.g.,--skip plan-design-reviewfor backend-only work) - Optional
--include-devex— add/li:plan-devex-reviewto the chain (default: skip) - Optional
--mode <full|minimal>—full(default) runs all 4 reviews;minimalruns office-hours + plan-eng-review only
Workflow
- Mode + skip decisions — confirm what's in the chain via AskUserQuestion (one question listing the proposed chain).
- Step 1: /li:office-hours — runs full skill. Output: design doc at
.claude/engineering/design-archive/<slug>-design.mdwith Status: APPROVED. - Auto-detect scope changes — if office-hours produced a design doc with major product-direction changes, ensure
/li:plan-ceo-reviewis in the chain (override --skip if needed; explicit operator override allowed). - Step 2: /li:plan-ceo-review — runs against the design doc. Output: CEO review log entry + verdict.
- Step 3: /li:plan-eng-review — runs against the design doc. Output: required Eng Review log entry + implementation task list + REPORT appended to design doc.
- Step 4: /li:plan-design-review (auto-detected: only fires if design doc has UI scope OR
--include-design-review) — UI/UX review log entry. - Step 5: /li:plan-devex-review (only fires if
--include-devex) — DX review log entry. - Synthesize: read all review-log entries from this run, render unified REVIEW REPORT in design doc.
- Final verdict: any review NOT CLEARED → autoplan exits "NOT READY"; all CLEARED → "READY TO IMPLEMENT."
Report format
Autoplan Status: <branch>
Chain: office-hours → plan-ceo-review → plan-eng-review → plan-design-review
Mode: full
Skipped: plan-devex-review (default)
Step 1/4 /li:office-hours: ✓ design doc APPROVED (.claude/engineering/design-archive/<slug>-design.md)
Step 2/4 /li:plan-ceo-review: ✓ SCOPE LOCKED (3 forcing-questions answered, 5 premises agreed)
Step 3/4 /li:plan-eng-review: ✓ ENG CLEARED (6 issues resolved, 2 critical gaps encoded as tasks)
Step 4/4 /li:plan-design-review: ⏸ SKIPPED — no UI scope detected
Final verdict: ✓ READY TO IMPLEMENT
Design doc: <path>
Task list: <N> implementation tasks
Next: begin Phase 1 implementation OR /li:ship (if work already done)
Compliance integration
- Each chained skill applies the active pack's compliance gates at invocation (
resolve_pack_field compliance.hooks; none in the neutral_defaultpack). - Autoplan aggregates: if ANY chained skill flagged a pack compliance gate, autoplan exits NOT READY.
Failure modes
/li:office-hoursreturns "NEEDS_CONTEXT" or "BLOCKED": chain pauses. Operator addresses, then re-runs autoplan (idempotent — reads existing design doc if present)./li:plan-ceo-reviewreturns REVISE: chain pauses. Operator updates design doc per CEO findings, then re-runs autoplan./li:plan-eng-reviewExit Plan Mode Gate fails: chain pauses. Operator fixes the design doc structure (REVIEW REPORT must be last h2).- Any chained skill times out: report which skill, allow operator to re-run that skill standalone, then resume autoplan.
Idempotency
Autoplan can be re-run safely. Each chained skill detects existing artifacts (design doc, review-log entries) and either:
- Updates them (if invoked with
--rerun) - Skips them (default — only fills missing steps)
Operator controls re-run granularity. Default = only fill what's missing.
Examples
Full pipeline:
> /li:autoplan
Chain: office-hours → ceo-review → eng-review → design-review
[runs all 4 in sequence]
✓ READY TO IMPLEMENT
Backend only, skip design-review:
> /li:autoplan --skip plan-design-review
Chain: office-hours → ceo-review → eng-review
[runs 3]
✓ READY TO IMPLEMENT
Minimal:
> /li:autoplan --mode minimal
Chain: office-hours → plan-eng-review (CEO + design + DX skipped)
[runs 2]
✓ READY TO IMPLEMENT
Blocked mid-chain:
> /li:autoplan
Step 2/4 /li:plan-ceo-review: ✗ REVISE — wedge specificity not established
Chain paused. Update design doc, re-run /li:autoplan when ready.
See also
/li:office-hours— design generator (first step of chain)/li:plan-ceo-review— step 2/li:plan-eng-review— step 3 (the required gate)/li:plan-design-review— step 4 (UI scope only)/li:plan-devex-review— opt-in step 5/li:ship— runs AFTER autoplan completes