One tool · mesh_mission
npm install -g @sym-bot/xmeshConnect from your tools · XMesh supplies the runtime. Your model and your subscription do the work.
A mission ends with more than an answer: whether it met your definition of done, tested by XMesh rather than taken on the agent’s word; an independent review, by default; and, when it cannot finish, the reason it stopped.
The ask can state what done means: a file that must exist, what it must say. When the agent reports it finished, XMesh tests those conditions itself and records the result beside the agent’s claim. An ask without them is left for a person to validate.
Work, review, research and assessment missions get a critic by default: it did not author the work and reviews it skeptically, and its verdict sits beside the doer’s own claim. It runs on the doer’s model unless you configure a second one; if it cannot run, the mission reads unreviewed.
A blocked mission says what it needs, and its root cause says what stopped it: a failure is a finding, not a silence.
The mission’s result, recorded checks and review stay readable by its id after the conversation that asked for it has ended.
This runtime is the shared substrate beneath XMesh: mission boundaries, collaboration, evidence, review, memory and a durable record around an agent that remains yours.