Set up a verified mission for high-risk work
For auth, billing, database migrations, or anything security-sensitive, a single ship verdict is not enough. Run the change as a mission: a builder implements it, an independent reviewer challenges it, and your repo's own checks run before the result is accepted. The output is a diff plus an Evidence Pack, not just a diff.
When to use a mission instead of an inline change
Use a mission when a mistake is expensive or hard to reverse: authentication and authorization, payments and billing, schema/data migrations, secret handling, or anything touching a critical hub. For routine edits, the inline verify loop is enough.
Create and run one
openthunder mission create # describe the goal
openthunder mission list # see your missions
openthunder mission run <id> # build → review → verify
The reviewer is independent from the builder by design , the point is an adversarial check, not a rubber stamp. When it finishes, inspect the evidence and the report:
openthunder mission verify <id> # check the evidence requirements held
openthunder mission report <id> # the mission report
A failed proof is a hard block: the change is not accepted just because a diff was produced.
From your editor or agent
- In Claude Code,
/ot-missioncreates a verified mission from the current context , use it for high-risk work instead of editing inline. - In VS Code, "OpenThunder: Create Mission from Selection" starts one from the code you have open.
- The app runs and tracks missions with a live Evidence panel showing the pack behind each verdict.
Let the router pick the models
When you create missions programmatically, omit the builder/reviewer fields and the router picks from available providers by task. To force a specific model, set the builder explicitly. OpenThunder is provider-neutral; agents are interchangeable and governed, never silently bundled.
Related
- Evidence Packs and provenance , what a mission produces.
- Decide if a change is safe to ship , the verdict a mission gates on.