Skip to content

fix(members): let a new crewmate start from a deleted one's memory - #22947

Closed
mbriones98 wants to merge 3 commits into
mainfrom
fix/22487-readopt-crewmate-memory
Closed

mbriones98 wants to merge 3 commits into
mainfrom
fix/22487-readopt-crewmate-memory

Conversation

@mbriones98

@mbriones98 mbriones98 commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: A crewmate created after another was deleted can start from the deleted crewmate's kept memory, and the delete confirmation says that the memory is kept.

When you delete a crewmate, its memory store is kept on disk, but nothing could attach that store to a crewmate again. Creating a crewmate always made a fresh store, and export, restore and import all need a live owner. The delete confirmation said the memory was kept but "you would have to set it up again", and no setup path existed.

Why it matters

Anyone who deletes a crewmate by mistake, or wants to rebuild one under a new name, loses everything it learned, even though the data is still there. The delete warning implied a recovery the product did not offer.

Not a goal

  • Copying, merging or moving memory between stores. The new crewmate binds to the kept store as it is.
  • Rewriting a store's identity, or adopting a store that has a live owner.
  • Loosening the live-owner requirement on export, restore or import.
  • A management screen for kept stores beyond the choice in the create dialog.

What changed (motivation → approach → change)

Symptom → cause: the persistence layer already accepts a member record bound to an existing store's owner_member_id, but every create path allocated a fresh store and never offered that binding.

Approach: let the create bind to a kept store's existing identity instead of allocating one. This copies nothing and keeps every existing guard. The alternatives (copying the store into a fresh one, or relaxing the import/restore owner check) would have rewritten identity or widened a guard.

Change:

  • Adopt module. The new member_memory_adopt module lists kept member stores that no crewmate owns. It also binds a new crewmate to one by name. A store is refused when a crewmate holds its id or binds it, and when an old-style crewmate (one with no member id) derives that id as its slug. It is also refused when the database identity is not exactly (owner_member_id, store), or when a kept chat-thread binding is keyed only by a former name. A name match never adopts. Adopting reuses the old crewmate's member id, so its kept chat thread, rules and activity follow the new crewmate too; this is the one exception to the rule that a new member id never reuses a kept store's owner id.
  • Locked write. persist_member_config re-checks under its lock that no crewmate written in the meantime binds the store, holds the id, or derives it as its slug.
  • API. POST /api/agents accepts adopt_memory_store, and GET /api/memory/ownerless-stores lists what is on offer (owner-gated like the other memory routes). If the create fails after the adopt, the store is left exactly as it was.
  • CLI. kirocrew agent create --adopt-memory-store <store>.
  • Create dialog. When kept stores exist, Advanced settings shows a Memory choice: start with an empty memory, or a deleted crewmate's kept memory. A pick is always sent as picked. If the server refuses it, nothing is created, the choice goes back to an empty memory, and the list is re-read.
  • Delete wording. The delete confirmation and the danger-zone note say the memory store is kept and a new crewmate can start from it later. The new text is in every shipped locale and en-XA.
  • Specs. memory-skills-hooks.md and learn-cron-dashboard.md describe the adopt path.

Backwards compatibility

Compatible: the new request field and CLI flag are optional and default to today's fresh allocation, and the new route is additive. No existing input is rejected.

Removes nothing: the diff only adds an optional create path and changes wording. Deleting a crewmate still keeps its store, as before.

Tests

  • test/test_member_memory_adopt.py (21 tests) covers:
    • re-adopting an ownerless store, through the API and the CLI;
    • refusing a store that has a live owner, and one that does not exist or whose record does not match;
    • refusing a store whose id an old-style crewmate derives as its slug, and one whose kept thread is keyed by a former name only;
    • the locked write refusing a concurrent adopt and a concurrently written old-style crewmate;
    • a create that fails after the adopt leaving the store ownerless and intact;
    • the listing offering only stores with no live owner.
  • NewCrewmateDialog.test.tsx checks that the Memory field shows only when kept stores exist and sends the picked store. It also checks that a refused pick resets to an empty memory and re-reads the list.
  • CrewEditorNextDialog.test.tsx checks the new delete wording.
  • Related suites pass: 1,705 backend tests (2 skipped) and 857 frontend tests in the members pages. Also clean: eslint, tsc, black, isort, flake8, mypy, the i18n check, the file line cap and docs-lint.

Manual verification

I captured the create dialog with the real component against stubbed /api responses (Vite + Playwright, dark and light). With one kept store on offer, the Memory field appears under Advanced, lists "Kept memory of Reviewer" beside "Start with an empty memory", and shows the pick with its hint. The backend path is covered by the tests above, not by a manual run.

Screenshots / video

Create dialog with the Memory list open (dark), and with a kept memory picked (light):

Memory list open, dark

Kept memory picked, light

The other theme of each state

Memory list open, light

Kept memory picked, dark

The delete confirmation only changes its sentence; its layout is unchanged. The English text now reads: "Delete crewmate {{name}}? New chats will no longer use it. Its files, memory and past conversations are kept, and a new crewmate can start from its memory later."

Related Issues

Fixes #22487

Pattern harvest

Rule candidate: review-prompt
Pattern: an identity or ownership check run on a config snapshot needs a twin inside the locked write that publishes it, and a UI that filters a stale pick out of the request should send it and let the server refuse it instead.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

@mbriones98
mbriones98 requested review from a team and KiroCrewCaptain as code owners October 21, 2026 23:18
@mbriones98
mbriones98 enabled auto-merge (squash) October 21, 2026 23:18
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 21, 2026
@github-actions

github-actions Bot commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: the change reuses a kept store's recorded identity under the existing lock and compare-and-set path, and leaves every live-owner guard as it was.

Notes

  • The description never mentions retire_stale_member_slot in members_thread.py. Called from _api_kirocrew_agents_create, it closes a live chat slot (DM slot) that is still pinned to the deleted crewmate's name. That changes chat-slot state during a create, so add it to "What changed".
  • The API create closes a stale live slot, but kirocrew agent create --adopt-memory-store does not. A CLI adopt made while a gateway still holds the deleted crewmate's thread open will have its thread open refused until that slot is closed. Either say so in the spec or route the CLI path through the same retire step.

[DESIGN-REVIEWED] ab17a57

Description read

[DESCRIPTION-READ] cbf2c4c7b6ed55a992ee0f26e5d5272b11cd18178ada34e301562629d9bc0d0f

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🔴 BLOCK

UX-level review of ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

UX-Verdict: BLOCK

The blind reader would not dare pick the kept memory, the PR's main control: the hint "nothing is copied" and the label "Kept memory of Reviewer" don't say what happens to that memory. No screenshot shows the refusal or the failed-load notice.

Blockers

Kept-memory choice reads as risky
Trigger: a user deletes a mate, opens New mate → Advanced, and sees the Memory field.
Outcome: the option "Kept memory of {{name}}" doesn't say the owner was deleted. The hint "…starts from that memory; nothing is copied." doesn't say whether the memory is moved, shared or used up. The reader said: "not sure I'd dare. 'Nothing is copied' confuses me: does it use the old memory or not?… does the Reviewer memory get used up?" and "I don't know who 'Reviewer' is". A user who can't predict the result keeps the empty memory, so the recovery this PR exists for goes unused.
Severity: every user who tries to recover a mate × the core value lost × every time.
Fix: change the label to "Memory of Reviewer (deleted mate)". Change the hint to say the result, e.g. "This mate takes over Reviewer's memory as it was. It won't be offered again."

Evidence gaps

  • The refused-pick state (create_memory_unavailable, with the field reset to an empty memory) appears in no screenshot. A screenshot of the dialog after a 409 would close it.
  • The failed-load notice for the list (create_options_failed naming "Memory") appears in no screenshot. A screenshot with /api/memory/ownerless-stores failing would close it.
  • The changed delete confirmation and danger-zone note appear in no screenshot. One screenshot each would close it.

Notes

  • The screenshots are older than HEAD's strings: they show "crewmate" where en.json now says "mate". Re-capture them against HEAD.
  • In shots 01 and 03 the open Memory list covers "Create crewmate". The list should open upward, or the dialog should keep the footer clear of it.
  • create_memory_unavailable says "It may already belong to another mate" even though the server returned a specific code. Name the cause, or drop that sentence.

[UX-REVIEWED] ab17a57

Description read

[DESCRIPTION-READ] e9b517473a117842a8797d796d3ca8c18dc3d7e4c7fcc3f242b8e326c49757e7

That sha256 names everything this verdict read: the PR description, plus the 7 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

Nothing blocks. One part of the change is not mentioned in the description, and one new check runs on every create, not just on adopts.

Notes

  • Item 6 — undeclared: when a crewmate adopts kept memory, the create now closes a live chat that is still pinned to the deleted crewmate's name (retire_stale_member_slot). The description never mentions this, though the HEAD commit title does ("open an adopted thread that is still live").
  • Item 5 — rides along: if create and config.agents[member].member_id: makes the locked write re-check old-style crewmates on every create, adopt or not. The description frames it as part of the adopt path ("re-checks under its lock that no crewmate written in the meantime binds the store"). Subtraction: run it only when adopting.

What this change ships

Inventory (9 items) — 7 justified

Intent: let a person rebuild a deleted crewmate by starting a new one from the memory it left behind, and make the delete warning say this is possible. This is a FIX (#22487).

  1. The create dialog's Advanced settings shows a Memory choice when kept stores exist; the default is still an empty memory — justified
  2. POST /api/agents accepts adopt_memory_store — justified
  3. New route GET /api/memory/ownerless-stores lists the kept stores — justified
  4. New CLI flag kirocrew agent create --adopt-memory-store — justified
  5. The locked config write refuses a create when a concurrently written old-style crewmate derives the same slug — rides along: runs on every create, not only adopts
  6. Adopting closes a live chat still pinned to the deleted crewmate's name — undeclared
  7. The delete confirmation and danger-zone note now say a new crewmate can start from the kept memory — justified
  8. The dialog's options notice now names a failed kept-memory read — justified
  9. Spec and feature-map rows describe the adopt path — justified

[FIRST-PRINCIPLES-REVIEWED] ab17a57

Description read

[DESCRIPTION-READ] 3f889741efc3fb0a159868700d8275313510d906bcd4d7a3b4c567331f08bf86

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 21, 2026
@github-actions

github-actions Bot commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — 🔴 changes requested (blocking)

GPT 6.1 found at least one blocking issue that must be resolved before merging ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff. 1 of 2 blocking finding(s) are security-class and were withheld from adjudication, so the blocking verdict stands.

This comment is updated in place on each push.

BLOCKING -- src/kiro_crew/dashboard/handlers/members_thread.py:134 -- Same-name adoption retains the deleted mate’s Trust grant

if slot is None or slot.mode != members_mod.DM_SLOT_MODE or slot.agent in ("", member):
Trigger: Delete a mate with an open, Trust-enabled DM, then recreate its name through POST /api/agents with its kept store and approval_mode="normal". This guard skips retirement; reopening and seed_for_turn preserve the slot’s already-set _member_approval_seeded and _trust.
Outcome: Tool requests that Normal mode would present for approval are instead auto-approved using the deleted mate’s grant.
Anchor: residual/security
Fix: Retire the existing adopted DM even when its agent name matches, so reopening seeds permissions from the new record.

BLOCKING -- src/kiro_crew/cli_commands_agent.py:283 -- CLI adoption leaves the retained DM pinned to the deleted mate

provision_or_adopt_member_memory(cfg, args.name, adopt_store)
Trigger: Delete a mate while its DM remains live in a running gateway, then adopt its store under another name using agent create --adopt-memory-store. CLI creation publishes the reused identity without retiring the old slot; api_member_thread finds that slot pinned to the deleted name.
Outcome: Opening the newly created mate’s chat returns 409 member_pin_mismatch until the obsolete slot is closed or the gateway restarts.
Anchor: defect
Fix: Reconcile the obsolete DM on thread opening after verifying the adopted member/store ownership, so recovery also works for CLI-created mates.

[BLOCK-MERGE] ab17a57
[GPT-REVIEWED] ab17a57

Adjudication (Opus 5.5) — is blocking on each finding proportionate?

F2 — MEDIUM harm. Opening the chat for a mate created from the CLI with an adopted store is refused with a 409 until someone closes the old slot or restarts the gateway. Nothing retries or recovers on its own.

  • The only place the stale slot gets retired is the API create path, src/kiro_crew/dashboard/agent_admin/crew_records.py:802-804. The CLI path at src/kiro_crew/cli_commands_agent.py:283 never calls it.
  • The 409 itself is at src/kiro_crew/dashboard/handlers/members_thread.py:276-283.
  • The real fix is cheap: retire_stale_member_slot (members_thread.py:112) already exists and only needs one more call site when the thread opens. The harm outweighs that cost.

F1 (fenced) — Deleting a mate does not close its live DM slot; crew_removal.py only stops deleting the store. Re-creating the same name with the adopted store gives the same slot key, and the guard at members_thread.py:134 skips the retire when slot.agent == member. seed_for_turn then returns early because the slot is already seeded (crewmate_approval.py:53), so the deleted mate's Trust grant stays in force. Deleting a mate and re-creating it under the same name is an ordinary sequence, not an extreme one, so this cannot be flagged.

[ADJUDICATION] ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff total=1 uphold=1 downgrade=0
UPHOLD F2 src/kiro_crew/cli_commands_agent.py:283 reason=harm-warrants-remedy
[GPT-ADJUDICATED] ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff

[ADJUDICATION-FENCED] ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff fenced=1 flagged=0
UPHOLD-FENCED F1 src/kiro_crew/dashboard/handlers/members_thread.py:134 -- Deleting a mate with a live Trust DM and re-creating the same name from its kept store is an ordinary sequence, and the same-name guard plus the already-seeded flag (crewmate_approval.py:53) carry the old Trust grant onto the new Normal-mode record.
[GPT-ADJUDICATED-FENCED] ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff

False positive or not applicable? A repository writer can comment:
/ai-review override gpt ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — 🔴 changes requested (blocking)

Reviewed ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff — this comment is updated in place on each push.

BLOCK: when a crewmate adopts kept memory through the new CLI flag, the deleted crewmate's open thread is never closed, so the new crewmate's thread refuses to open while the gateway runs. This PR closes that thread only on the dashboard create path.

BLOCKING — src/kiro_crew/cli_commands_agent.py:283
provision_or_adopt_member_memory(cfg, args.name, adopt_store)
Trigger: A crewmate "Ana" is deleted from the dashboard while its chat thread is still open in the running gateway (the delete route does not close that thread). The user then runs kirocrew agent create Bea --adopt-memory-store <Ana's store>. Bea takes over Ana's member_id and store, so the open thread's key is now Bea's (member_slot_key(agent.member_id, agent.memory_store)). The CLI path never calls retire_stale_member_slot, which the dashboard create calls at crew_records.py:802-804.
Outcome: Opening Bea's thread reaches members_thread.py:405 (elif slot.agent != member_name:) and returns 409 member_pin_mismatch ("the thread is pinned to a crew the registry no longer names"). Sending is refused the same way (chat_send.py:845). This lasts until the old thread is closed or the gateway restarts.
Anchor: defect
Fix: Put the stale-thread close where both create paths go through it. One option: in the thread-open path, when the binding now resolves to the crewmate being opened, close the thread still pinned to the old name instead of refusing. Another: have the CLI tell the running gateway to close it after an adopting create.

[BLOCK-MERGE] ab17a57
[OPUS-REVIEWED] ab17a57

Verdict parsed from the review's SHA-scoped output markers for commit ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff.

False positive or not applicable? A repository writer can comment:
/ai-review override fable ab17a579d5ceabc17e8ccfe52a7bba349bcb04ff false-positive: <reason, 20 to 500 characters>

@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 22, 2026
Matt Briones and others added 3 commits November 4, 2026 03:18
Deleting a crewmate keeps its memory store, but every route into a
member store resolves it through its live owner, so the kept store
could no longer be opened, exported or restored, and nothing could give
it an owner again.

A create can now name a kept store that no crewmate owns
(adopt_memory_store on POST /api/agents, --adopt-memory-store on
`kirocrew agent create`). The new crewmate is bound to the store's own
recorded member id: nothing is copied and the database identity is
unchanged; only the record's display owner follows the new name. It is
refused for a store that has an owner, an undeclared or default store,
a database whose identity does not match its record, an id a legacy
crewmate derives as its slug, and a kept chat-thread binding keyed by a
former name only. A name match never adopts, so a same-named create
still starts fresh, and export, restore and import keep requiring a
live owner. GET /api/memory/ownerless-stores lists the stores on offer
(owner-gated), and the create dialog offers them under Advanced
settings when there are any.

The delete pane and its confirmation now say the memory is kept and
that a new crewmate can start from it later, in every shipped locale.

Fixes #22487
…p a stale pick visible

The create's locked write now refuses a crewmate written meanwhile whose
name derives the adopted member id as its slug. The create dialog sends a
kept-memory pick as picked, so a pick that went stale is refused by the
server, and after that refusal the pick returns to an empty memory.
… kept memory fails to load

A crewmate that adopts a kept store derives the deleted crewmate's thread key.
A slot still live there kept the deleted name and the open refused it (409
member_pin_mismatch). The create now closes that slot the way the tab close
does, so the next open rehydrates it from history under the new crewmate.

The New mate dialog says a failed kept-memory read in its options notice
instead of offering nothing, and the feature map names the Memory choice and
GET /api/memory/ownerless-stores.

Part of #22487
@iamwhatever
iamwhatever force-pushed the fix/22487-readopt-crewmate-memory branch from b46c0d7 to ab17a57 Compare November 4, 2026 03:32
@iamwhatever

Copy link
Copy Markdown
Contributor

Rebased on main and addressed the GPT 6.1 findings on b46c0d7 in ab17a57:

  • Live slot pinned to the deleted crewmate (members_thread / member_memory_adopt): an adopt-create now closes a live DM slot on the adopted key whose agent is someone else, through the same close path as the tab close (close_slot: saved to history as closed, slot leaves the live set). The next POST /api/members/{slug}/thread rehydrates it under the binding, which names the new crewmate. New test test_a_live_thread_of_the_deleted_crewmate_opens_as_the_new_one drives create → live slot → delete → adopt → open: 409 without the change, 200 with slot.agent == 'scout'.
  • Memory-list failure hidden (NewCrewmateDialog): the kept-memory query's error now feeds the dialog's options ErrorNotice, naming the Memory list. Test added.
  • Feature map: the Crew Members row now names the Memory choice, handlers/memory_admin.py / member_memory_adopt.py, adopt_memory_store and GET /api/memory/ownerless-stores.

Rebase note: main renamed crewmate to mate in UI copy and dropped Built from and the second Create button. The delete strings keep main's wording in every locale; English adds that a new mate can start from the kept memory. Also fixed the two comment-history hits.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Nov 4, 2026
@bolichen97

Copy link
Copy Markdown
Contributor

[operator: bolichen97] Continuing this in #33148; thanks for the start.

bolichen97 added a commit that referenced this pull request Nov 8, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 8, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 8, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 10, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 11, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 11, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 13, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 13, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 13, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
@mbriones98

Copy link
Copy Markdown
Contributor Author

Closing: bolichen97 is continuing this work in #33148.

Written by the Crew Tasks Bot on behalf of mbriones98.

@mbriones98 mbriones98 closed this Nov 13, 2026
auto-merge was automatically disabled November 13, 2026 18:51

Pull request was closed

@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Nov 13, 2026
bolichen97 added a commit that referenced this pull request Nov 13, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 17, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 19, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
bolichen97 added a commit that referenced this pull request Nov 19, 2026
Deleting a bot keeps its memory store on disk, but every route into a
member store resolves it through its live owner, so the kept store could
not be opened, exported, restored or picked again, while the delete text
read as if the memory were safe.

A new bot created from the dashboard can now take over a kept store
that no bot owns: the New bot dialog offers it under Advanced settings,
POST /api/agents binds the new bot to the store's recorded member id,
and GET /api/memory/ownerless-stores lists what is on offer. Nothing is
copied. The create clears every approval grant on that thread's session
and closes the deleted bot's thread still live on that key, so its pin
and grants do not carry over, and the delete text says a new bot can
start from the memory.

Builds on #22947 by @mbriones98.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

merge conflict Branch has merge conflicts with its base — author must resolve before merge

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Deleting a crewmate orphans its memory store with no route back

3 participants