What is the goal of create-spec?

To elicit requirements and generate a SPEC.md task specification via interactive questions.

How many options do I present per question?

2-4 concrete options plus an 'other' to capture edge cases.

What if I need to adjust the SPEC after generation?

Refine requirements with additional questions and regenerate or update the SPEC.md content.

create-spec

Scanned

npx machina-cli add skill srobinson/helioy-plugins/create-spec --openclaw

Files (1)

SKILL.md

1.0 KB

Create Spec

Process

1. Explore First

Before asking questions, understand the codebase:

ls -la
git log --oneline -10
find . -type f -name "*.ts" -o -name "*.py" -o -name "*.sh" | head -30

2. Elicit Requirements

Ask ONE question at a time
Offer 2-4 concrete options (plus "other")
Confirm understanding: "So [paraphrase], correct?"
Focus on WHAT, not HOW

3. Generate SPEC.md

# Task: <name>

## Goal

<clear statement of objective>

## Success Criteria

- [ ] <verifiable outcome 1>
- [ ] <verifiable outcome 2>

## Constraints

- <limitation 1>

## Notes

<implementation guidance if needed>

Good vs Bad Criteria

"All existing tests pass" (good)
"New endpoint returns 200 for valid input" (good)
"Code is clean" (bad — subjective)
"Performance is good" (bad — unmeasurable)

Source

git clone https://github.com/srobinson/helioy-plugins/blob/main/plugins/helioy-tools/skills/create-spec/SKILL.mdView on GitHub

Overview

This skill guides you to craft a clear Task SPEC.md by eliciting requirements one question at a time. It helps define WHAT to build before autonomous execution starts, reducing ambiguity and aligning expectations.

How This Skill Works

First, it explores the project context with lightweight checks. It then asks one targeted question at a time, offering 2-4 concrete options plus 'other' and confirming understanding with paraphrase. It then generates a SPEC.md template with sections for Goal, Success Criteria, Constraints, and Notes.

When to Use It

Starting a new feature or tool
Before building an autonomous agent or automation
When stakeholders disagree on project scope or requirements
During mid-project requirement refinement
When documenting tasks for handoff or archival

Quick Start

Step 1: Initiate the create-spec process to start exploration
Step 2: Answer with one question at a time and select from 2-4 concrete options plus 'other'
Step 3: Review the generated SPEC.md template and adjust as needed

Best Practices

Start with a concise project snapshot to anchor scope
Ask one question at a time to avoid overload
Offer 2-4 concrete options plus an 'other' to capture edge cases
Paraphrase the user's answer to confirm understanding
Keep the SPEC.md objective, measurable, and future-proof

Example Use Cases

Define a data ingestion task with clear success criteria
Outline a chatbot feature and its interaction flow
Specify a deployment automation task with constraints
Clarify an API integration between services
Describe a UI component enhancement with measurable outcomes

Frequently Asked Questions

Add this skill to your agents