Workspace AI lock: block every agent/assistant and hide all AI interaction

Chiusa
  • Qualcosa non va?
#614 1 commento 0 reazioni 1 assegnatario Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

@PelikanFix16 ci sta già lavorando.

Dal 18/10/2026.

Valutazione

Questa issue non è ancora stata valutata.

Descrizione

agent:planning enhancement

Founder direction (Hubert, supervising session, 2026-10-18). A workspace can be completely free of AI: no agent or assistant may read it or act in it, including a member’s own personal assistant. Workspace settings must provide the switch; the Agents tab and other AI interaction surfaces disappear while blocked. This is a server-enforced workspace policy.

Milestone: 2 — complete application and release (brief). Extends the foundation’s optional AI/human collaboration and F-016/F-018/F-022/F-027. Persona: a team that wants ordinary human collaboration with AI entirely excluded from this workspace.

Acceptance criteria

  • AC-1 — Workspace control. An authorized workspace administrator can persistently disable/enable AI in workspace settings, on computer and phone. Members without settings authority cannot change it. Reload/reconnect preserves the policy; existing-workspace migration/default and the current state are explicit.
  • AC-2 — Total supported AI access denial. When disabled, all supported agent/assistant entry points are denied both reads and actions: personal assistants, embedded agents, external MCP clients, API-provider runs and official-CLI runtimes, automatic/scheduled/proactive work, bootstrap/orientation, tools and subscriptions. Owner/admin membership, an existing connection, token or standing grant never overrides the block. No blocked workspace content or metadata is supplied indirectly through global discovery/search, context assembly, files/citations, event feeds or caches. Enforce this on the server across all workspace projects, private and DM contexts.
  • AC-3 — Immediate cutover. Disabling fences active/queued runs and existing sessions/subscriptions across API/worker/runtime processes: no subsequent read, model dispatch, streamed delivery, write or result commit may cross the disabled policy. Cancel/stop in-flight work safely; a late provider response cannot publish. Re-enable requires current authority and never silently replays work queued while blocked. Record/test the precise commit/cancellation boundary; no stale-cache grace window.
  • AC-4 — Human-only interface. Hide the Agents tab, personal-assistant entry points, AI buttons/menus/shortcuts, suggestions and AI onboarding within that workspace. Existing tabs/deep links receive a clear blocked state without launching AI. Preserve the settings switch and normal human conversation, tasks, maps, wiki, files, search and collaboration. Preserve existing native work/history; disabling does not delete data. Other workspaces retain their own AI policy and operation.
  • AC-5 — Observable verification. Docker API/worker/runtime/MCP tests cover two owners, administrator/member/agent identities, already-issued grants/tokens, active/queued/scheduled jobs, revoke/reconnect, cross-workspace context and a late in-flight result. Include actual multi-process enforcement and provider-call negative controls. Chromium/WebKit desktop/phone tests verify settings authorization, hidden AI surfaces, direct links, reload and working human collaboration; independently evaluate the final candidate with the bounded review policy.

Ownership and interfaces: single implementation owner @PelikanFix16; independent evaluator @Zamojski5. Coordinate existing #68 personal-assistant runs, #152/#160 external co-work, #153 scheduling, #281 runtime activation, workspace access/settings and Agents UI. These are shared interfaces, not permission exceptions; the workspace block takes precedence. Keep this one coherent outcome, without child-issue fragmentation.

Contract status / next action: founder outcome accepted; queued, implementation not started. Record the workspace policy/cutover contract in product/development docs and obtain one bounded peer assessment before implementation. Existing WIP #612/#613 continues. No founder approval queue. No matching workspace-wide AI exclusion issue was found in the existing backlog.

Lingua principale
TypeScript
Stelle
3
Fork
0
Merge medio
17h 51m
PR unite (30g)
367

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di ColdPhase/flux

Tutte le issue di ColdPhase/flux

Issue simili

Altre issue su TypeScript

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.