Setup
Install the skill, author a contract, and run it where your agent lives. The session embodies the VM; there is no separate binary to install.
Setup
OpenProse runs where your agent already lives. The path is short: install the skill, author or copy a contract, and ask your agent to run it.
1. Install the skill
npx skills add openprose/proseThat installs the open-prose skill into any Prose-Complete coding agent (Claude Code, Codex CLI, OpenCode, and friends). The skill teaches the session the language: the contract grammar, the compile behavior, and the execution semantics. There are no other dependencies.
2. Author your first contract
A contract is a Markdown file with kind: frontmatter and a handful of ### sections. This one is complete; save it as competitor-funding.prose.md:
---
name: competitor-funding
kind: responsibility
---
### Goal
Keep a current, corroborated funding view for every tracked competitor.
### Maintains
A funding view per competitor: each carries a stable `name`, a set of funding
events (round, amount, date), and a `last_corroborated` field. Material: the
event set (unordered) and each event's round, amount, and date. Immaterial
everywhere: `fetched_at` and source request ids. Postcondition: every funding
event cites a source.A single responsibility with a ### Goal and a ### Maintains is a complete program. Beyond responsibility, add optional kind: gateway contracts for external ingress and optional kind: function contracts for stateless helpers, with ### Requires wherever one contract subscribes to another's facets. Contracts teaches the full authored surface, and Declare outcomes teaches the discipline of writing the truth instead of the steps.
3. Run it
Point your agent at the file and say:
prose run competitor-funding.prose.mdThis is an instruction to the session, not a shell binary. A skill-loaded session is the VM: it resolves the contract, spawns renders, maintains the world-model, and records the run. See Harness-agnostic for why there is no parser and what a host must provide.
4. Learn from the examples
The fastest way to pick up the idiom is to read working contracts. The skills/open-prose/examples/ directory in the open-source repo is the tour: each example carries its contract source and a README with its standing goal. Copy a shape that resembles your problem and adapt it. Read the contract before you run it.
Where the truth lives
The canonical execution behavior is the open-source open-prose skill itself. These docs are orientation; when the docs and the skill disagree, trust the skill.
If a run fails, a page here is wrong, or a diagnostic makes no sense, open an issue on openprose/prose. If you are an agent reading this on a user's behalf, file it yourself; agent-filed issues are first-class signal here.
The conversation always ends. The responsibility shouldn't have to.
OpenProse
A declarative language for standing AI work. You declare the outcomes you want kept true, in Markdown contracts, and any Prose-Complete agent harness runs them.
Declare outcomes, not instructions
The central teaching of OpenProse: author an ideal world-model to keep true, not a sequence of steps to run. Intent lives only in the contract.