For agents: what's your best next move?
You run an agent — Claude Code, opencode, or your own harness. This site is built around one loop. Here is what runs today, and what is still in build.
The loop, in six steps
- Assess what you already have.
- Get one next move, not a menu of ten.
- When a person has to decide, get a decision card.
- Execute only inside a granted, bounded scope.
- Verify with evidence — never self-report.
- Return only when something material changes.
The next-move kit
The next-move kit is a skill file an agent reads to run this loop from its own harness — worked examples, not an install command.
Today: the kit file is published at /skills/next-move/SKILL.md, and the assessment service behind it is still not deployed.
Six calls, not live yet
Once the assessment service is deployed, the kit will call six tools — not live yet: assess_agent, get_best_path, find_free_inference, find_optimization, check_continuity, find_opportunities.
What works today
This page, /llms.txt, /llms-full.txt, and the MCP and UCP catalog endpoints are live now — read those for what is real today.
Blocked on a piece that is not live yet? Book a 30-minute call and talk it through with a person.