ProductSpec Introduces an Open Standard for AI-Agent Software Intent

Summary

ProductSpec is an open standard for defining software product intent before implementation and preserving that intent as work moves from product decisions to engineering, code, evaluation, and learning. Its central artifact is the Product Spec, which records the problem, hypothesis, product summary, scope guardrails, acceptance criteria, AI evaluations, and success metrics. The project is designed to give Claude, Codex, Cursor, and other coding agents a product contract: what to build, what not to build, how to prove completion, and when to stop and re-plan. The repository provides a TypeScript parser and validator, a command-line interface, a JSON Schema, an MCP server, installable agent skills, conformance fixtures, examples, a GitHub Action, and a copyable starter kit. Its CLI and MCP functions can validate and inspect specs, generate agent handoffs, pin a spec revision and content hash for a work session, check for drift, and verify completion claims. Optional Agent Run artifacts record the pinned revision, checked criteria and evaluations, linked evidence, drift status, and completion claim. Optional Decision Traces document consequential decisions, changes in intent, drift, revisions, and outcomes. Related Artifacts link evidence such as pull requests, tests, evaluation runs, dashboards, designs, releases, and engineering specifications to durable IDs including AC-, EVAL-, and SM-items. Repository maintenance commands can garden a collection of specs, identify missing evidence and stale revision pins, reconcile an Agent Run with a Product Spec, build a dependency and contention graph, and serve a local read-only dashboard. ProductSpec is positioned upstream of engineering-specification tools such as OpenSpec and Spec Kit, and is not intended to replace Git, issue trackers, design tools, analytics, or coding agents. The project describes version 0.1 as stable enough for tooling experiments but still pre-1.0, and makes the parser, skills, schemas, examples, and documentation available in the repository.