Two new options in hub.yaml / hub.config.ts that address real problems we hit daily: agents making claims about external state without checking, and agents ignoring team memory.

Fact Checker

New workflow.fact_checker option. When enabled, the generated orchestrator prompt includes a Mandatory Verification section that forbids the agent from stating the status of any external resource without first checking it with the appropriate tool.

This covers:

• PR status (merged, open, closed, approved)
• Branch state (ahead, behind, conflicts, existence)
• Deploy status (deployed, failed, in progress)
• CI/CD pipeline results
• Issue/task status
• Service health

export default defineConfig({
  workflow: {
    fact_checker: true,
    pipeline: [...]
  }
});

If the agent can’t verify something, it must explicitly say so instead of guessing.

Memory Enforcement

New memory.enforce option. When enabled, the Team Memory section changes from a soft recommendation to a hard rule — the agent must run search_memories at the start of every interaction.

export default defineConfig({
  memory: {
    path: "./memories",
    enforce: true
  }
});

The enforced prompt instructs the agent to:

1. Extract key topics from the user’s message
2. Search memories with relevant keywords before doing anything else
3. Factor found memories into the response
4. Save new learnings after completing work

This ensures institutional knowledge (past decisions, conventions, gotchas) is always considered, not just when the agent feels like it.

Upgrade

hub update