Skip to content

ROADMAP.md

Latest commit

 

History

History
153 lines (119 loc) · 4.7 KB

ROADMAP.md

File metadata and controls

153 lines (119 loc) · 4.7 KB

AIR Prompt Runtime Roadmap

This roadmap is for the prompt-version AIR repository.

It does not cover the private backend implementation except where prompt-side compatibility or public documentation requires mention.

Roadmap principles

  • keep the prompt-version repo honest about scope
  • strengthen AIR as a prompt-native runtime
  • preserve map-first, active-step-only execution
  • add agent-bridge capability without turning AIR into the agent runtime
  • improve usability before increasing complexity

Current status

v0.1.0 — Initial public prompt runtime

Status: released

Delivered:

  • AIR Core Runtime Prompt
  • AIR Control Surface Prompt
  • AIR Default Starter Profile
  • AIR Handoff Card Template
  • README for prompt-version usage
  • orientation-first startup
  • project execution map
  • active-step-only artifact discipline
  • handoff continuity
  • tested behavior on ChatGPT 5.4, Claude Sonnet 4.6, and Gemini Thinking

Near-term roadmap

v0.1.1 — Prompt/runtime polish

Goal: Tighten the public prompt kit and improve first-use clarity.

Planned:

  • improve README and quick-start clarity
  • add cleaner visual assets for repo and posts
  • standardize repo description and topics
  • refine release notes and public-facing language
  • document Gemini Thinking boot note clearly

Success criteria:

  • new users can boot AIR from the README without extra explanation
  • repo clearly reads as prompt-version only
  • no ambiguity between public prompt kit and private backend

v0.2.0 — Agent Execution Bridge

Goal: Allow prompt-version AIR to generate structured execution specs for external agent runtimes like OpenClaw.

Planned:

  • define AGENT_EXECUTION_SPEC
  • add rules for when AIR may emit agent-ready instructions
  • keep agent execution external to AIR
  • add documentation and examples
  • add operator guidance for AIR -> agent -> AIR loop

Success criteria:

  • AIR can produce one clean, active-step-specific agent execution spec
  • the spec is concrete enough for an external agent to act on
  • AIR remains the planning/control layer, not the execution runtime

v0.3.0 — Stronger continuation and governance

Goal: Make continuation, orbit discipline, and state handoff more robust.

Planned:

  • refine handoff card semantics
  • make orbit model more visible in docs
  • improve restoration flow for continuation sessions
  • add guidance for multi-session project progression
  • document conflict handling and degraded state more explicitly

Success criteria:

  • continuation works reliably with less operator correction
  • handoff cards feel like real restoration objects
  • orbit model is understandable to new users

Mid-term roadmap

v0.4.0 — Specialized prompt profiles

Goal: Add specialized public prompt profiles for distinct project classes while keeping the default starter simple.

Potential additions:

  • technical / architecture profile
  • research synthesis profile
  • GTM / narrative profile
  • compliance / evidence-heavy profile

Rules:

  • keep starter profile as the default
  • do not fragment the public repo into prompt sprawl
  • specialized profiles must remain compatible with the same AIR runtime laws

Success criteria:

  • specialized profiles improve fit without breaking the common runtime model
  • users can still understand the default path quickly

v0.5.0 — Public evaluation and examples

Goal: Show how AIR behaves in practice across real project scenarios.

Planned:

  • example project boots
  • example handoff flow
  • example active-step artifact progression
  • example agent execution bridge output
  • comparison between generic chat workflow vs AIR runtime workflow

Success criteria:

  • users can see AIR in action without guessing how it works
  • examples reinforce the actual workflow rather than abstract theory

Longer-term direction

Backend-connected AIR

This remains outside the public prompt-only repo for now.

Longer-term goals may include:

  • backend compilation of AIR artifacts
  • stronger typed state and validation
  • automated project map updates
  • direct agent orchestration
  • runtime persistence beyond prompt-only continuity

Public prompt repo role in that future:

  • remain the public prompt-native layer
  • remain useful even without backend access
  • stay honest about what is prompt-only vs backend-backed

Non-goals for the prompt-version repo

These are not current goals for this repository:

  • publishing an SDK
  • pretending the prompt kit is a backend runtime
  • turning AIR into an autonomous agent executor
  • bloating the repo with too many profiles too early
  • replacing the private backend repo with markdown abstractions

Working model

AIR should continue to operate as:

  1. project activation
  2. roadmap creation
  3. active-step artifact generation
  4. optional agent execution bridge
  5. state handoff
  6. continuation