manage teamsYour management experience — scope control, review processes, decision tracking — encoded as a protocol AI can follow.
No routes, no frameworks, no dependencies. Just what the product is.
actorsWho uses it? Define the people, not the code.
entitiesWhat data does it store? The product data model.
featuresWhat can it do? Atomic product capabilities.
storiesHow do things interact? The product narrative.
permissionsWho can do what? Role-to-feature matrix.
deniedWhat must it NOT have? The most important field.
A complete product boundary in one JSON file. Human-scannable in minutes.
{ "name": "chat-system", "version": "1.0.0", "description": "Real-time chat with group conversations and read receipts", "author": "kim", "actors": ["Admin", "Guest", "Member"], "features": ["createGroup", "listConversations", "readReceipts", "sendMessage"], "denied": { "Reaction": "Keep messaging simple, no emoji reactions", "editMessage": "Messages are immutable once sent", "voiceCall": "Text-based communication only" } }
One generates, one verifies. Human makes the final call.
Scans the lock, approves or denies. Never reads code. Reads the lock in minutes.
Writes code, generates the lock from the codebase. Submits for review.
Independently verifies code against the lock. Does NOT trust the Worker.
Guides for humans and AI. Each document serves a specific role.
The full specification. File format, fields, validation rules, lifecycle, and conventions.
Read spec → GuideFor AI Workers. 10-step process to analyze a codebase and generate a product.lock.json.
Read guide → GuideFor AI Reviewers. Field-by-field verification process. How to catch violations.
Read guide → SpecificationLock defines WHAT. Plan defines HOW. Implementation blueprint format.
Read spec → ToolQuantify product complexity from a lock. PLS formula validated against 10 open-source products.
Read scoring → ToolTrack changes, decisions, and scope across sessions. Context reconstruction for AI.
Read worklog →