Build Product Specs your agents can actually execute.
ProductSpec.io is a privacy-friendly editor and Product Harness for AI-native software work.
Draft product intent, publish clean Markdown to GitHub, and give coding agents a build contract they can execute against using the open source ProductSpec standard.
New to model evaluation? See AI eval examples.
Drafts are not saved on our servers. They stay in your browser until you publish to GitHub.
Open existing Product Specs from GitHub.
Sign in with GitHub to open Product Specs from a repo.
Add only the sections that sharpen the document.
Agent Handoff and Evidence
Turn this Product Spec into a build contract, then track the proof an agent or team should return before the work is called done.
Agent Handoff
A plain-English build contract for Claude, Codex, Cursor, or an engineer. Copy it into an agent, a GitHub Issue, or a pull request description when you are ready to implement.
# Agent Handoff: Untitled Spec ## Build Contract Implement ProductSpec revision 1. Satisfy every acceptance criterion before claiming the work is done. Stay inside scope. If implementation pressure changes product intent, stop and propose a Product Spec update or Decision Trace. ## Product Summary Not specified yet. ## Scope Guardrails - Not specified yet. ## Must Satisfy - AC-1: Add acceptance criteria before implementation. ## Evidence The Agent Must Produce - Return proof for each acceptance criterion before claiming the work is done. ## Evidence To Return - Pull request URL - Verification result for each acceptance criterion - Screenshots or demo link if UI changed - Decision Trace if implementation changes product intent
Evidence
ProductSpec.io has not verified this work yet. These are the proof slots an agent or team should fill before claiming the Product Spec is complete.
Acceptance criterion not written yet.
Success metric not written yet.
Use Decision Trace only when implementation changes product intent.