In everyday use, work through Codex CLI in the brain repository or through your own application. Give that agent or system an explicit workflow for refreshing sources, retrieving context, and checking evidence.
After updating your notes
serenity -C ~/my-brain sync
serenity -C ~/my-brain search "release decision"Sync brings in changed sources. Search checks that the material you expect is available. If you configured extraction, run it separately:
serenity -C ~/my-brain extractBefore starting a task
serenity -C ~/my-brain search "project constraints"
serenity -C ~/my-brain ask "What did I decide about this project?"ask needs composition models. Check citations before carrying an answer into another tool or a consequential decision.
Check structured actions
For a brain with active constraint precepts, check actions against those recorded constraints:
serenity check --actions '[{"action":"spend_over","params":{"amount":500}}]' --jsonThe amount here is example input, not a default spending policy. The result depends on the precepts in your brain.
| Exit | Interpretation |
|---|---|
| 0 | Pass or no applicable constraints. Inspect the verdict. |
| 2 | A recorded constraint was violated. |
| 1 | Unverified or an error. |
No applicable constraints does not mean an action is safe. Free-text plans currently return unverified because a classification provider is not wired into that CLI path.
When something feels wrong
serenity -C ~/my-brain doctor
serenity -C ~/my-brain status
serenity -C ~/my-brain connectors statusRead the status before repeating a large ingest. Start troubleshooting with one source and a query whose answer you already know.