MEP MEEP-ReadOnly-v1

Released 2026-04-29

Read-only peer-agent context surface.

Added

spec/meep-readonly-v1.md New sibling spec defining a small, public-but-unlinked Markdown page that external peer agents (Grok, ChatGPT, Gemini) fetch on session start. Unidirectional, agent-scoped, pull-on-demand. Sized to act as effective system-prompt context without blowing token budget.

First instance: Hastings (Leo) at nukasoft.ai/leo. Page owner: Rita. Source scaffolded at site/_pages/leo.md.

Required sections

Stable Fundamentals, Current Strategic Priorities, Open Questions, Recent Decisions, Active Workstream State, Sync Notes, Changelog.

Token budget Soft cap ~1,500 words / ~2,000 tokens per page.

Design Decisions

Not a Standing Standup. The standup (v2.1) is bidirectional and project-scoped. MEEP-ReadOnly-v1 is unidirectional and agent-scoped. Different problem, different primitive.
No time-based pruning in v1. A 48-hour prune was considered and rejected: peer agents that check in weekly would silently lose history they never read. Manual curation by the page owner instead. v2 path is last-read telemetry.
No scheduled rebuild timer in v1. External peer agents do not need 10-minute freshness; the timer adds operational overhead with little signal value. Triggers are manual + a light EOL hook that queues drafts but does not auto-publish.
Security via obscurity + content discipline. noindex,nofollow, robots.txt disallow, no sitemap entry. Content rule: if a competitor reading this would be a problem, it does not go on the page. Auth tokens are v2 backlog.
URL shortener / redirector deferred to v2. Direct nukasoft.ai/{agent-name} is the v1 primitive; an aka.ms-style redirector to an off-repo canonical surface is the expected next step.

All MEP releases

Current specification