skills.rest
Simon-YHKim
Simon-YHKimCOMMUNITY
Explore all skills by this author →
GitHub

autoplan

Runs CEO, design, engineering, and DX plan reviews sequentially with auto-decisions.

Updated Apr 5, 2026
Search Tags:#design review#plan review#plan mode#engineering review#ceo review#review pipeline#auto-decision
One-click install
npx skills add https://github.com/Simon-YHKim/eject-button --skill autoplan-simon-yhkim
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: autoplan
Source: https://github.com/Simon-YHKim/eject-button/tree/main/.claude/skills/autoplan
Command: npx skills add https://github.com/Simon-YHKim/eject-button --skill autoplan-simon-yhkim

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Running a full multi-perspective review of a plan normally means invoking four separate review skills and answering 15-30 intermediate questions. This Skill runs the entire review gauntlet automatically, making decisions on your behalf using 6 explicit decision principles and only surfacing genuine taste decisions at a final approval gate. ## Core Features & Use Cases - Sequential Review Pipeline: Reads the CEO, design, engineering, and DX review skills from disk and executes them in strict order at full depth, with phase-transition summaries between each. - 6 Decision Principles: Auto-answers intermediate questions using completeness-first, boil-the-lake, pragmatic, DRY, explicit-over-clever, and bias-toward-action rules, with an audit trail logging every decision. - Dual-Model Review: Runs both a Claude subagent and Codex as independent outside voices per phase, building consensus tables and flagging disagreements. - Use Case: You have a rough plan file for a new feature. Invoke the auto-review pipeline and get back a fully reviewed plan with premises confirmed, scope decisions made, architecture validated, and a short list of taste decisions awaiting your approval. ## Quick Start Ask the assistant to run the autoplan auto-review pipeline on your current plan file and make the intermediate decisions for you.

Frequently Asked Questions about autoplan

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run all plan reviews automatically without answering questions?▼

Invoke the autoplan skill on your plan file. It loads the CEO, design, eng, and DX review skills from disk and runs them sequentially, auto-deciding intermediate questions using 6 decision principles and surfacing only taste decisions at a final approval gate.

What decision principles does the auto-review pipeline use?▼

It uses 6 principles: choose completeness, boil lakes within the blast radius, be pragmatic, reject duplication per DRY, prefer explicit over clever code, and bias toward action. Conflicts are resolved by phase-specific tiebreakers.

Which decisions are never auto-decided in the review pipeline?▼

Premises about what problem to solve always require human confirmation, and User Challenges where both models agree the user's stated direction should change are never auto-decided. These go to the final approval gate with full context.

Does the auto-review pipeline run design and DX reviews every time?▼

No. Design review runs only when UI scope is detected (2+ view or rendering terms in the plan), and DX review runs only when developer-facing scope is detected, such as API, CLI, SDK, or agent-related terms.

Can I recover my original plan after running the auto-review?▼

Yes. Phase 0 captures a restore point by writing the plan file's full contents to a timestamped file under ~/.gstack/projects/ before any changes, and prepends a restore path comment to the plan file.

Related Skills

Explore More Product & Management Skills→
TwoTonedTwoToned

gstack-openclaw-office-hours

Convert vague product ideas into formal YC-style design documents.

Official
Advanced
ceedaragentsceedaragents
794

implementation

Implements requested code changes following existing patterns and verifies them with tests.

Official
Basic
Sma1lboySma1lboy
136

interrogate

Spawns multiple LLM reviewers to adversarially review code changes and synthesizes a verdict.

Community
Advanced
prabaljainnprabaljainn

grill-me

Interviews the user question-by-question to stress-test plans and design decisions.

Community
Basic