clawplays/ospec

Spec-driven, agentic workflow framework for AI coding agents. Turn a request into a verifiable goal loop — plan, act, verify — with durable specs and evidence in your repo. Works with Claude Code, Codex, Gemini, OpenCode, and plain CLI.

What it solves

AI coding assistants often store critical requirements and decision-making context only in chat history, making it difficult to inspect, review, and resume work across different sessions or tools. OSpec solves this by introducing a spec-driven development (SDD) layer that persists the entire workflow—proposals, designs, plans, and verification evidence—directly in the project repository.

How it works

OSpec implements a "Loop Engineering" framework (plan $\rightarrow$ act $\rightarrow$ verify) that integrates with AI agents like Claude Code, Codex, and Gemini. It operates through two primary modes:

  1. Change Flow: A lightweight path for simple requirements (init $\rightarrow$ change $\rightarrow$ verify/finalize) that creates a sequence of markdown files (proposal, tasks, verification, review) to track progress.
  2. Goal Workflow: A more rigorous agentic process for complex work. It uses a controller layer to derive a task graph, dispatch parallel worker agents, enforce reviewer gates, and require deterministic verification evidence before a goal is marked complete.

The system uses a "Protocol Shell" (including .skillrc and a .ospec/ directory) to maintain project knowledge and state, allowing any compatible AI agent to pick up exactly where a previous one left off.

Who it’s for

Developers using AI coding agents (such as Claude Code, Codex, Gemini, or OpenCode) who want a verifiable, audit-able, and persistent record of AI-driven changes rather than relying on ephemeral chat logs.

Highlights

  • Repository-Based Context: Moves requirements and plans from chat history into version-controlled files.
  • Agentic Controller: Supports parallel worker dispatch and independent reviewer gates for high-risk tasks.
  • Tool Agnostic: Works across various AI harnesses and CLI workflows.
  • Verifiable Loops: Ensures work is proven through verification evidence and independent reviews before archiving.