Execute multi-agent AI programs defined in prose files.
npx skills add https://github.com/mistakenot/auto-stack --skill open-prose-mistakenot
Please help me install this Agent Skill. Skill: open-prose Source: https://github.com/mistakenot/auto-stack/tree/main/.agents/skills/open-prose Command: npx skills add https://github.com/mistakenot/auto-stack --skill open-prose-mistakenot
OpenProse formalizes and operationalizes multi-agent AI sessions so you can write, validate, and run reproducible workflows instead of juggling ad-hoc prompts. It solves brittle coordination, hidden state, and replayability problems by defining a manifest-driven wiring phase and a VM-driven execution phase with durable state and clear contracts.
Ask the assistant to run the hello-world example by saying: prose run examples/01-hello-world.md
High-intent search queries and answers about installing and using this skill.
Multi-agent AI workflows are orchestrated reproducibly by defining them in prose files, which a manifest-driven wiring phase resolves into an executable manifest for a VM to run with durable state and clear contracts.
Manifest-based two-phase execution separates agent workflow processing into a wiring phase that auto-resolves services and emits a manifest, followed by a VM execution phase that spawns subagents and logs append-only state for resumption and auditing.
To run a multi-agent program, you ask the assistant to execute the prose file, such as by saying 'prose run examples/01-hello-world.md', which triggers the VM to spawn subagents and manage state via copy-on-return.
Yes, multi-agent workflow state backends support SQLite and PostgreSQL, alongside filesystem and in-context options, allowing you to persist execution state and enable resumption for long-lived orchestrators.
Multi-agent AI sessions lose state and fail to replay when lacking a durable execution model, a problem solved by using append-only state logging and dependency resolution via .deps and prose.lock to enforce reproducible runs.
The best way to inspect completed multi-agent runs for cost analysis is using retrospective inspection tools that analyze the append-only state logs and profiler outputs generated during the VM-driven execution phase.
AIMFllysEdits StudySolo workflow canvas nodes and edges with dry-run validation.
ezagent42Analyze milestone execution data to generate structured retrospective recommendations.
LerianStudioAutomate subagent-driven development with fresh subagents per task and automated reviews.
managedcodeDevelop agentic .NET applications with Microsoft Agent Framework orchestration.
