Redaction on write and the credential vault

The idea

Secret-shaped values are removed before events, steps, tool results and model context are stored, and customer credentials are sealed with AES-256-GCM bound to org and credential.

The problem
Tokens pasted into work or tools leak into logs and prompts.
Why it matters
Stored records and model inputs hold no readable secrets.
Where it stands
In production as of

A step-by-step replay of this idea is not published yet. The library adds one when a recorded run shows it.

  • Secrets used by reference

    Passwords, codes and captured tokens are stored as references; the platform fills or stages them into the guest and masks output, so no value reaches events, logs or model context.

  • Stated data retention

    Settings shows what is kept and for how long from the constants in force; full model inputs are kept 7 days and only the requester can read or delete them.

  • Gate on all outgoing text

    Commits, PR text, comments, messages and reports go through rule-based checks (length, structure, banned phrases, attribution, secrets) and a style check against examples a person approved. Failing text gets a limited number of rewrites.

  • Operator setup, console and restore-proven backups

    A proven platform operator stores vendor apps and keys in the product. Values are sealed and the environment wins. The operator sees every org without SQL, and a weekly job proves the newest backup restores.

Turn the numbers into shipped work.

Crewshal runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.