Guide · Agent skill
How to use the Feature Forge agent skill: install and run it
Updated · written with AI from the project's own documentation and checked against our catalogue data
Feature Forge is an agent skill that turns a rough feature idea into a formal requirements specification before any code is written. It is for product managers who need a spec before engineering starts, and for developers who want edge cases and non-functional needs pinned down first. After this guide you will know how to install the skill, how to trigger it with a normal request, and what output to expect.
Feature ForgeJeffallan/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the feature-forge ZIP from the skill page.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into .claude/skills/ in your project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
- 4Confirm your agent supports skills. Check that your agent supports SKILL.md skills and, ideally, the AskUserQuestions tool.
- 5Make a natural request. Ask for a feature definition, requirements, user stories or acceptance criteria in plain words.
- 6Answer the interview. Respond to the structured and open questions about users, goals, security and edge cases.
- 7Review the spec file. Check the generated specs/{feature_name}.spec.md before engineering starts.
What is an agent skill, and what does Feature Forge make the agent do?
An agent skill is a folder containing a SKILL.md file and optional reference files that tells an AI agent how to handle a specific kind of task. When the skill is installed and relevant to your request, the agent follows its instructions.
Feature Forge makes the agent act as a requirements analyst. It works from two perspectives: a PM view focused on user value, business goals and success metrics, and a Dev view focused on technical feasibility, security, performance and edge cases. It runs a structured interview, then writes a specification with EARS-format functional requirements, Given/When/Then acceptance criteria, non-functional requirements, an error-handling table and an implementation TODO checklist. The skill saves the result as a file named specs/{feature_name}.spec.md.
Install the Feature Forge skill
Download the ZIP for feature-forge from the skill page. Then choose one of the following options.
For a personal install across projects, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Single project: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Check what the skill needs before you start
Feature Forge needs an agent that supports SKILL.md skills. It refers to the AskUserQuestions tool for structured choices. If that tool is unavailable, the agent falls back to plain questions.
Task subagents are optional. They are used for multi-domain features and are described in references/pre-discovery-subagents.md.
The skill does not run scripts and does not use the network.
- Required: an agent that supports SKILL.md skills
- Referenced: AskUserQuestions tool (plain-question fallback)
- Optional: Task subagents for multi-domain features
- Runs scripts: no. Uses network: no.
Trigger the skill with a natural request
You do not need a special command. Describe what you want to define or specify, using words from the skill's trigger list such as requirements, specification, user stories, acceptance criteria, PRD or EARS. The skill's description says to use it when defining new features, gathering requirements or writing specifications.
Expect questions before any document. The skill is built to interview you first and will ask for clarification on vague input, so have your target users and goals ready.
- "Help me define requirements for a CSV export feature."
- "Write user stories and acceptance criteria for team invitations."
- "Create an EARS-format spec for password reset."
Answer the interview and review the output
The interview runs in stages: discovery of the goal, users and value; interview from both PM and Dev angles; writing EARS requirements; validating acceptance criteria with the stakeholder, with key trade-offs presented as choices; and then an implementation checklist.
The output follows a template: overview and user value, functional requirements, non-functional requirements, acceptance criteria, error handling table and implementation TODO checklist. Review each part before handing the spec to engineering.
- Discover: goal, target users, user value
- Interview: PM and Dev questions, including security and edge cases
- Document: EARS requirements
- Validate: acceptance criteria and trade-offs
- Plan: implementation checklist
Safety notes and tips
Because the skill runs no scripts and uses no network access, its effect is limited to the conversation and the files the agent writes, such as the spec file. Still, read any spec before sharing it, since it reflects your answers.
For best results, give concrete answers about users and constraints. The skill is designed to reject vague requirements such as "make it fast", so state measurable targets where you can. Load the reference files only when you need them; the skill is designed to pull in EARS syntax, interview questions and acceptance criteria formats as the context calls for.
- Skip it for one-line bug fixes or small tweaks.
- Skip it when requirements are settled and only code needs writing.
- Expect it to need answers; it will not produce a finished spec from a one-line idea alone.
What people use it for
- Writing a full spec with EARS requirements for a new user-facing feature before development.
- Turning a vague product idea into user stories with testable Given/When/Then acceptance criteria.
- Listing edge cases, error handling and non-functional needs such as security and performance before building.
- Producing a consistent PRD-style document for a team that reuses one format.
- Creating an implementation TODO checklist from a finished requirements set.
Frequently asked questions
How do I install the Feature Forge skill?
Download the feature-forge ZIP from the skill page and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ in a project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
What does Feature Forge require?
It requires an agent that supports SKILL.md skills. It references the AskUserQuestions tool and falls back to plain questions if that tool is unavailable. Task subagents are optional for multi-domain features.
Is Feature Forge free to use?
The skill is released under the MIT licence. The source does not state any price, so check the skill page and your agent provider for any costs.
Does Feature Forge run scripts or access the network?
According to the skill's listing, it runs no scripts and uses no network access. It works through the conversation and writes the spec file.
Where does the spec file get saved?
The skill saves the specification as specs/{feature_name}.spec.md, where the feature name comes from your feature.
Can I use it for a small bug fix?
It is not recommended. The skill is designed for full requirements interviews, which would be overkill for quick one-line fixes or small tweaks.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Feature Forge suits product managers and developers who need a structured, testable spec before building a new feature, and teams that want a consistent PRD or user-story format. Skip it for small fixes, for features whose requirements are already settled, or if you expect a finished spec without answering questions.
Download the skill