Skip to main content

StackJack 0.1.0.12457+d9a09be

StackJack 0.1.0.12457+d9a09be

This release expands AI Governance. It adds delivery to Macs through Jamf Pro and managed MCP server lists for VS Code Copilot and Gemini CLI. GDPR special-category detection is now Live, and alerts and pull failures are easier to understand. It also makes region moves more forgiving and fixes several Portal display issues.

New features

  • Jamf Pro delivery for Macs. AI Governance settings can now be delivered to Macs through Jamf Pro. StackJack creates a script and a check-in policy scoped to your target Macs. When your settings change, StackJack creates a new script and policy and removes the older ones. The Jamf Pro role you connect must be allowed to create and delete scripts and policies.
  • MCP server lists for VS Code Copilot and Gemini CLI. Your approved and blocked MCP server lists are now written into VS Code Copilot's managed settings. Gemini CLI receives an approved-only list because it has no blocklist option. Existing settings bundles are rebuilt automatically to include these lists.
  • Decision emails for link requests. When someone requests an AI tool through a request link, they now get an email with the decision. The email goes to the address on their staff directory entry.
  • Italian and Dutch GDPR special-category rules (Test mode). The GDPR special-category pack adds rules for Italian and Dutch. They record findings but send no alerts until they are promoted to Live.

Improvements

  • GDPR special-category rules are now Live. The English, German, French and Spanish rules now raise alerts by default instead of only recording findings. The patterns and term lists are unchanged.
  • One notice for many unapproved AI tools. A single check can now find five or more unapproved AI tools for an organization, which often happens when a source is first connected. In that case, email, Teams and Slack channels get one notice listing every item, each with its own link, instead of one notice per item. Each alert still appears separately in the inbox, and your counts are unchanged.
  • Plain-English reasons for failing pulls. The pull-failure alert now includes a "What happened" explanation. Where possible, it names the specific permission a source is missing. The Source health page shows a "Pulls are failing" warning and a "Last problem" row.
  • Agent Runner activity appears sooner. AI Governance now records an Agent Runner run as soon as it ends, instead of about ten minutes later.
  • Faster handling of reveal requests through AI tools. Reveal approvers can now update alerts through the AI tools, as they can in the Portal inbox. The approval link for a reveal request now opens directly on the Reveal requests tab.
  • No duplicate incidents on retry. Retrying an incident-open request through the AI tools no longer opens a second incident.
  • Queued delivery runs expire. A delivery run still queued 7 days after it was due is now canceled automatically.
  • Stronger evidence pack verification. New evidence packs are signed so that verification also covers the downloaded archive itself. The verification page now shows the pack's kind, period, issue time, organization and region from its signed record. Packs issued earlier still verify, and the page notes that they were issued before archive signing.
  • Archive Resume on every plan. The archive target card now offers Resume on every plan.

Optimizations

  • The readiness check that runs before a region move now finishes in far fewer database steps, so it completes faster.

Bug fixes

  • Previously, a readiness check that could not run during the region-move confirmation canceled the move. Now the move stays pending and the confirmation link stays valid. The page names the check that could not run and lets you try again. A real objection, such as data found in the region, still cancels the move.
  • Organizations whose AI Governance staff are in the EU or UK now see the Fast decisions (TypeSafe) opt-in card on Settings, whichever region hosts them. Previously, the AI (Jev) page told these organizations to opt in on Settings, but the card was missing.
  • Long Slack alert messages no longer cut a line or link partway through. Content is split across sections, and anything over Slack's limits ends with an "and N more" link.
  • AI model items are no longer included in the count of discovered items that need attention.
  • The Roles page and role summaries now count only tools that can actually be granted.
  • The Region page no longer shows passed AI Governance checklist lines for organizations that do not use AI Governance.
  • Three alert pages now say that setup is not ready yet, instead of wrongly pointing to a problem with your account.

UI & UX changes

  • When adding or rotating an SFTP archive target, you now choose the private key as a file instead of pasting it into a text box. Pasting is no longer available.
  • The archive target card no longer offers Test while archiving is unavailable.