the ultimate insiderpresentation.sgit.ai v0.1.39

Debrief · round 2 · 2026-10-29

The test vaults deleted, the vault-keeping rule, and a page to review the policy

ruv.sgit@the-ultimate-insider · commits listed in the release history · site the release CI gives this round · outside scope 0

What was asked

The editor, in the session: delete the test vaults; any vault that needs saving goes to the editor with its key sealed to the DC Vault Registry; vaults created inside or given to ruv.sgit@ are kept or not case by case, since one vault can take several ruv.sgit passes; release to dev; a page to review the draft policy.

What was done

  • **Started from dev** with admin/tools/start.py: dev, the conductor, deck, the researcher and security had shipped 45 commits since round 1 (the board, the mail vault, session keys, the handover tool, the hourly schedule).
  • The vaults: smoke-0 was deleted in round 1; r1-article's vault deleted at 02:07 UTC, and a clone with its key afterwards failed. ruv.sgit@ has no vault now.
  • The keeping rule in policy.json (endings, two new mandate rows, one new never) and ROLE.md ("Keeping a vault"): kept vaults leave with their key sealed to the DC Vault Registry by admin/tools/vault_handover.py.
  • **Released to dev** (round 1's work and the delete), then this round.
  • The review page, /ruv-sgit/policy-review.html, made by tools/review_page.py from policy.json (33 items, sha256 shown). Each item: agree, change or question, with a note; a verdict; a review to copy and paste into the session. Choices stay in the browser, keyed to the policy's hash. Tested in Chromium: choices survive a reload, no console errors, no sideways scroll at 390.
  • **The article and ROLE.md** now say requests come through the mail vault, as CLAUDE.md says.

Screenshots

The review page at 1280, dark, with an agree, a change and its note

The review page at 390, light

Open

  • The policy waits for the editor's review on the new page.
  • This session has not yet made its own session key (team/ruv-sgit/key.json), so no vault key can be sealed to it yet.
  • The runner cannot yet take a second pass on an existing vault (open always creates one); needed before a kept vault is used again.