Skip to content

feat: run a chat in a Kiro Cloud Sessions sandbox - #21260

Merged
iamwhatever merged 1 commit into
feat/session-environments-gitfrom
feat/session-environments-kiro-cloud
Oct 21, 2026
Merged

iamwhatever merged 1 commit into
feat/session-environments-gitfrom
feat/session-environments-kiro-cloud

Conversation

@kyleseaman

@kyleseaman kyleseaman commented Oct 16, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Let a chat run entirely in a Kiro Cloud Sessions sandbox, chosen from the same environment chip as a Coder workspace.

Session Environments (#19833) moves a chat's files and shells to a Coder workspace, but the agent itself always runs on the gateway. A user who already has Kiro Cloud Sessions cannot hand a chat to that sandbox from Kiro Crew; kiro-cli's own --cloud flag is the only way in.

Why it matters

A chat can borrow Kiro's managed cloud sandbox, with its repositories checked out there, without anyone running Coder infrastructure. Because that sandbox sits outside Kiro Crew's tool gate, the operator has to opt in deliberately, and an enterprise policy has to be able to keep it off.

Not a goal

  • Running Kiro Crew's tools, MCP servers, agent spec or PreToolUse gate inside the cloud sandbox: the sandbox runs Kiro's own default agent.
  • Files, Git and terminal views for a cloud session: those stay Coder-only. On a cloud session they show a neutral notice saying only the sandbox's own agent reaches its files.
  • Moving a session that already started, locally or on Coder, into the cloud.
  • Provisioning Kiro Cloud Sessions access or a KAS backend for the user.
  • Other providers (Fly.io Sprites, boxd, AWS Fargate): each is its own sibling PR on the same base.

What changed (motivation → approach → change)

A Session Environments profile can say "run this whole chat in Kiro's cloud". Kiro Crew does not build a sandbox. It asks KAS for one with the same _meta.kiro keys kiro-cli's --cloud flag sends, confirmed against the installed kiro-cli 2.21.1's ACP client (newSession / loadSession). It refuses to start, rather than run locally, when KAS does not offer a cloud-sandbox target.

Part What it does
src/kiro_crew/config, config-baseline.json, src/kiro_crew/dashboard (handlers/core.py) session_environments.kiro_cloud_profiles: at most 20 {id, name, repositories} records, each with at most 10 owner/name repositories. A save that gives one id to both a Coder and a Kiro Cloud profile is refused with 400 profile_id_collision, writing nothing
src/kiro_crew/platform (governance_parts/catalog.py), src/kiro_crew/coder (session_launch.py), src/kiro_crew/session_allocation_parts (get_or_create.py) capabilities.kiro_cloud_sessions, opt-in. When the policy or the session's profile sets ANY other scope, a cloud start needs that row enabled in the POLICY; a profile can withdraw it, never grant it. The profile is the one the session's tool calls get: app-bound first (the chat slot's own app, which session allocation sets on the provider as owning_app; never transcript metadata), then agent, then surface. An unreadable ceiling refuses. Every decision writes a SEL governance event
src/kiro_crew/coder (session_launch.py, session_browse.py) the choice row records its provider; prepare_for_session returns PreparedCloudSession; the cloud id is recorded only if the same provider and profile are still chosen, else the start fails and the runtime is killed. Browse raises CloudSandboxOnly before any Coder call
src/kiro_crew/acp (harness/kas_cloud.py, harness/base.py, runtime.py, session_handle.py, session_handle_turn.py, session_handle_lifecycle.py, session_handle_runtime_contract.py, session_provider.py), src/kiro_crew/providers (acp_runtime_start.py, acp.py, base.py), src/kiro_crew/session_lifecycle_parts (fleet.py) KasCloudSandboxHarness sends executionTarget: cloud-sandbox plus the repositories (or isEmptyWorkspace), no MCP servers and no custom agent; every other host's request is unchanged (session_request seam, harness-parity H13). Replayed approvals are held by a resume hold only the provider's cloud start takes (hold_cloud_resume; the runtime's load_session is unchanged), plus two reader seams (holds_replayed_permission, unroutable_permission) every other host answers as a no-op, so the runtime's shared start and resume paths never test for the cloud harness. The cloud-only branches are the cloud harness itself, the provider's cloud start, and the handle's drain and teardown, which read is_cloud_sandbox. The keep-running rules below
src/kiro_crew/dashboard (chat_runner.py, chat_turn/turn_spec_hooks.py, handlers/session_environments.py), src/kiro_crew/context_assembly (turn.py) a failed reopen leaves a transcript error row (meta.kind = cloud_session_replayed); the environment endpoint returns the recorded provider; browse and file routes answer 409 cloud_sandbox; the [PROJECT] line names the sandbox
website/src/pages, website/src/components, website/src/api, website/src/app-sdk, website/src/i18n a Kiro Cloud Sessions section on the Settings card after Coder's three steps, with its own buttons, setup steps and what opting in gives up; the chip names a cloud profile by its repositories, or "Empty workspace" when it has none; the Files panel shows the same info notice for a recorded cloud chat (sending no request) while a browse the gateway refuses with 409 cloud_sandbox is a rejected request and renders as an ErrorNotice with Ask the agent (no tree, Refresh or terminal), like any other listing failure; the error card renders cloud_session_replayed in the reader's language. Strings in all 12 catalogs
src/kiro_crew/change_card_catalog.py, docs kiro_cloud_profiles and default_profile are code-exec. Specs: providers.md, acp-client.md, session.md, config.md, governance.md, docs/feature-map/README.md

Backwards compatibility

Compatible: kiro_cloud_profiles is a new optional key, empty by default, so every existing profile, chat and save behaves as before. Every new runtime branch (no drain cancel, held approvals, local-only teardown, the startup notice) is keyed on the cloud harness, so other hosts keep their wire and teardown. One change-card classification moves: editing session_environments.default_profile shows the code-exec card; the values it accepts are unchanged.

Readers that now meet a kiro_cloud choice row, unchanged by this PR, each staying closed:

  • Reader: session_environment.session_runs_remotely answers True, so local-only tools stay refused for a cloud chat, as for a Coder one.
  • Reader: coder/retention.py:warn_unsaved_work asks session_browse.unsaved_work, whose _resolve raises CloudSandboxOnly; the check logs it and the archive goes on with no notice.
  • Reader: providers/acp_lifecycle.py:_bound_remotely answers True, so a non-runtime backend refuses with REMOTE_NEEDS_KAS instead of starting locally.
  • Reader: the setting.change card for session_environments.default_profile now asks for code-exec approval.

Removes nothing: no key, route, tool or setting goes away, and the session_request seam returns its inputs unchanged for every existing host, and the two reader seams do nothing on them.

Cloud session lifecycle

The cloud agent keeps running when Kiro Crew lets go of it, so Kiro Crew never cancels or deletes it on its own:

When What happens
Gateway stops (graceful or not) The local KAS relay ends. A cloud turn is not cancelled: the shutdown drain and provider shutdown skip any provider whose turn_survives_disconnect is true. A user's Stop still cancels
Chat reopens session/load with the recorded id and sessionSource: remote; no second sandbox
An approval was pending KAS replays it during the load. It is held (60 s at most) until the session registers, then reaches the chat's normal approval path on the next turn; past that it takes the audited reject path
A new sandbox can't be recorded The start fails and deletes that sandbox first: it holds no conversation and nothing else points at it
A session/new no handle was returned for Deleted, because nothing records its id: a start whose local setup refuses (permission routing, mode, snapshot) after session/new answered, a late session/new answer nobody adopts, and the entitlement probe's own session. A session/load never counts, so a failed reopen keeps its session
Chat ends An ordinary chat unregisters locally and leaves the cloud session in place. Only an incognito or temporary chat deletes its cloud session. Deleting or archiving an ordinary chat keeps it too: the session is the user's own record in Kiro (listed in Kiro Web and the IDE), it may hold uncommitted work that exists nowhere else, and _kiro/session/delete cannot be undone, so removing a Kiro Crew row is not taken as consent to destroy it. Sessions Kiro Crew no longer points at are listed and deleted in Kiro Web (app.kiro.dev) or the IDE's cloud-session menu (Delete Session), per Kiro's docs
Reopen fails A new sandbox starts with the conversation carried over; the old session is left in Kiro's cloud, where it may still hold work. The transcript error says: "Couldn't reopen the cloud session, so a new sandbox started with this conversation carried over. Files and changes from the earlier sandbox aren't in it. The earlier session is still listed in Kiro Web (app.kiro.dev)."
Subagent continued after its parent's row aged out It keeps its own placement: the Kiro Cloud profile, provider and recorded cloud session, so it reopens its sandbox instead of being re-recorded as Coder.

A chat's environment locks only once Kiro's cloud holds the session; a start refused before then leaves the chat free to move back to this machine. Riders: default_profile moves from cannot-widen to code-exec, because a default can now bind every new chat to a sandbox outside the tool gate; Kiro Cloud Sessions joins the i18n glossary.

Stack

PR Base Adds
#19833 feat/session-environments-coder main Coder environments
#21256 feat/session-environments-git #19833 per-session git proxy and protected-branch push rule
#21259 feat/session-environments-health #19833 workspace health card
#21257 feat/session-environments-sprites-boxd #21256 Fly.io Sprites and boxd
#21258 feat/session-environments-fargate #21256 AWS Fargate
this PR, feat/session-environments-kiro-cloud #21256 Kiro Cloud Sessions
feat/session-environments-git-panel #21256 the remote Git panel

Merge order: #19833 first; then #21256 and #21259; then Sprites/boxd, Fargate, this PR and the Git panel PR, each on #21256. This PR adds nothing to #21256's git proxy, because a cloud session has no git proxy. It carries no Git panel: whichever of this PR and the Git panel PR lands second adds the panel's cloud placement (a neutral notice for a recorded cloud chat, an error for a refused request) and its test, reusing this PR's cloudSandbox string.

Tests

  • test/test_kiro_cloud_session.py (new, 104 tests against a KAS stub): the session/new / session/load shapes, the empty workspace, and the refusal when KAS offers no cloud-sandbox. Governance: a ceiling on any scope refuses until the policy opts in; a profile cannot grant; an unreadable ceiling refuses; every decision is audited. Lock: a refused start can move back; a cleared, Coder or other-cloud choice during the reply fails the start and kills the runtime. Keep-running: the drain and close_all send no cancel or delete for a cloud turn while an ordinary KAS turn still cancels; the next start reloads the same id; a replayed approval reaches the normal approval event, also with a stranded request queued behind it, and an unregistered one expires; every served host answers the reader seams as a no-op, the shared destroy and turn drain carry no getattr probe for is_cloud_sandbox, the runtime's load_session takes no hold, and the provider's cloud start does; only incognito/temporary chats delete; a refused cloud start, an unadopted late session/new and the entitlement probe delete the session they minted; a failed reload after a successful load deletes nothing; the fallback notice is logged and surfaces once, including through a real dashboard turn (test/turn_harness.py's run_turn; removing the call in _run_chat fails it).
  • test/test_session_environments_config.py, test/test_change_card_setting_classes.py, test/test_session_environments_start_gate.py: round-trip and collision refusal, default_profile is code-exec, a cloud-only config counts as configured. test/test_acp_bindings.py, test/test_acp_harness_contract.py, test/test_acp_runtime_load_session.py, test/test_harness_parity.py follow the new property and the moved _load_session body.
  • Frontend: website/src/test/SessionEnvironmentsCard.test.tsx (the cloud section sits outside Coder's steps, with its own labels and plain-view cost), website/src/test/RemoteFilesPanel.test.tsx (a recorded cloud placement renders an info notice with no browse call; a 409 cloud_sandbox browse is an alert with Ask the agent and no Refresh, tree or terminal, and a real listing failure stays an alert with Ask the agent), website/src/test/SessionEnvironmentChip.test.tsx, website/src/test/ErrorCard.test.tsx (the replay error is localized), website/src/test/settingsCoverage.test.ts.
  • Run on this head: the eight backend files above, the repo gate tests (spawn audit, stdlib shadow, wait-discipline ratchet, channel boot hot reload, config module boundaries, five-whys Windows lock, session allocation, subagent parts, sections facade, gateway runtime, Slack gateway), the session-handle owner bindings, and every test that names the drain, session_lifecycle_parts, the harness base, the answer task, load_session or the session-environment handlers, plus the remote git and remote session suites: on a5e034a313, 6939 passed (every test naming the session handle, terminate_session, the cloud placement or the session-environment launch, browse and handlers, plus the gate and binding suites) and 30769 (every test naming the governance catalog, the chat runner and turn hooks, allocation, the fleet drain, change cards or config sections). Of the rest, test_markdown_agent_specs::test_the_strict_listing_raises_where_the_glob_reads_an_empty_directory, test_acp_refactor_facade::test_every_facade_patch_the_scan_cannot_resolve_is_a_listed_one, test_tool_call_classification_provenance_bound::test_a_payload_the_encoder_could_not_render_is_refused_not_retained_unmeasured, test_harness_parity::test_added_line_gate_reports_without_enforcing and test_backend_cards_advertised::test_no_new_runtime_site_gates_a_field_backed_line_on_a_set fail identically on this PR's base; test_dashboard_server_composition_contract's boot-chain test failed only in a worktree carrying old built app files and passes on a fresh checkout of this head; test_slot_create_default_agent passed on its own rerun. Reverting either relocated cloud hunk (the replayed-approval hold in the turn owner, the private-chat delete in the lifecycle owner), or discarding a held replay's id on its first drain read, or skipping the cloud start's resume hold, or not marking a minted session unclaimed (refused start, late answer, probe), makes its test fail. The five frontend files, 156 passed. tsc (npm run typecheck), i18n:check, eslint on changed TS/TSX, flake8 and isort on changed Python, mypy on the 27 changed source modules, the black gate and scripts/docs_lint.py pass.
  • Not run here: the full CI suites (CI runs them), native Windows and macOS, guest Python 3.9, and live provider infrastructure. No real KAS sign-in with Kiro Cloud Sessions was used: KAS is a stub, so whether the cloud keeps the turn across the relay's SIGTERM is taken from Kiro's docs, not measured.

Manual verification

Not verified against a real cloud account. Whether a cloud turn actually keeps running after the local KAS relay stops has not been tested against a real Kiro Cloud Sessions account. Every test here ran against local KAS stubs; the keep-running behaviour above is what Kiro Crew sends and stops sending, and the claim that the cloud keeps the turn comes from Kiro's docs.

KAS limits, as found:

Limit Value Source
Concurrent cloud sessions 10 Kiro's Cloud sessions page
Region us-east-1 only same page
Attach-stream re-attach budget 60 s (backoff 250 ms to 10 s) constant in KAS's client code
Provisioning deadline 25 min constant in KAS's client code
Idle timeout not documented, not measured none
Maximum lifetime not documented, not measured none
  • Rendered the Settings card, chip picker, Files panel and transcript error row from this head with stubbed API answers (below).
  • Still to do: one live cloud session through a KAS sign-in that offers cloud-sandbox, including a gateway restart mid-turn, a reopen with a pending approval and a session/load resume.

Screenshots / video

Captured from this head's components in a Vite dev server on 127.0.0.1, with every /api call answered from fixtures (no gateway, no KAS).

Settings, the Kiro Cloud Sessions section after Coder's steps: what opting in gives up, setup steps open (step 1 names the backend as the Agent Harness picker shows it, "KAS (kiro-agent)"), the profile expanded with a pending edit, so the sticky save bar shows, and its own Remove and Add buttons:

Kiro Cloud Sessions settings section, dark

Files panel. Left: a recorded cloud chat gets an info notice and the panel sends no browse request. Right: a browse request the gateway refuses with 409 cloud_sandbox is a rejected request, so it is an error in its own words ("Couldn't open this chat's files: …") with Ask the agent and no Refresh, tree or terminal:

Files panel on a cloud chat: info notice and refused-request error

Transcript error after a failed reopen (the error card shows the localized sentence):

Transcript error: a new sandbox started; the earlier session is still in Kiro Web

Default for new chats, open: a Kiro Cloud profile can be the environment every new chat starts in:

Default-profile select listing a Kiro Cloud profile

Composer environment chip: a Coder profile shows its directory, a Kiro Cloud profile its repositories, and an empty one "Kiro cloud sandbox":

Environment picker with Kiro Cloud profiles

Light theme and full card

Kiro Cloud Sessions settings section, light

Whole Session Environments card, light

Whole Session Environments card, dark

Related Issues

Part of #10433

no linked issue: this PR closes nothing on purpose; #10433 tracks the whole Session Environments stack and stays open until it lands.

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

@kyleseaman
kyleseaman requested review from a team and KiroCrewCaptain as code owners October 16, 2026 14:38
@kyleseaman
kyleseaman requested review from mbriones98 and removed request for a team October 16, 2026 14:38
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 16, 2026
@github-actions

github-actions Bot commented Oct 16, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 4c925456c6f85f4a6524d1ec28bd9c164add7b61 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Before merge, confirm that a real KAS session started with _meta.kiro.executionTarget: cloud-sandbox comes up in the sandbox and reopens with sessionSource: remote. The keys were copied from kiro-cli 2.21.1's client, and the stub only echoes them back.

[DESIGN-REVIEWED] 4c92545

Description read

[DESCRIPTION-READ] 3f232dfb4047728e14e5e5489d2454d462fd4adb5da7dcf394e6b7dd1670e5df

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 16, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — 🟡 CONCERNS

Premise-level review of 4c925456c6f85f4a6524d1ec28bd9c164add7b61 — 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/CONCERNS are advisory.

First-Principles-Verdict: CONCERNS

Preserving replayed approvals across a reopen ships five pieces of machinery beyond the Goal, when the existing audited reject already handles the replay safely.

Not justified as shipped

  • Item 7 — oversized: the Goal is "Let a chat run entirely in a Kiro Cloud Sessions sandbox". Holding approvals that KAS replays goes further than that. Without the hold, a replay takes _answer_unroutable_permission, which fails closed and is audited. To avoid that, the PR adds CloudResumeHold and hold_cloud_resume, two new HarnessAdapter seams (holds_replayed_permission and unroutable_permission, each overridden by one class and called at one site in runtime.py), a new reader branch, and _cloud_replayed_permissions with the _cloud_kept path in session_handle_turn.py.
    Subtraction: delete all of these, so a replayed approval gets the existing reject and the user asks again.
    Clears when: the mechanism is removed, or a human amends the Goal.
  • Item 5 — symptom-level: choice rows already record provider (CHOICE_PROVIDERS). The collision only happens because profile_provider works out the provider from the id alone.
    Subtraction: look profiles up by (recorded provider, id), and delete ProfileIdCollision, check_profile_ids_unique, _refuse_shared_ids, ambiguous_profile_ids and the 400 profile_id_collision. That is 14 call or definition sites in src/.
    Clears when: the lookup takes the provider into account and the collision code is deleted.
  • Item 8 — one consumer, generalized: LLMProvider.startup_notice: tuple[str, str] is a generic slot, but acp_runtime_start.py:986 sets the only value it ever holds and _surface_startup_notice is its only reader.
    Subtraction: replace it with a single boolean, cloud_session_replayed.
    Clears when: a second notice producer exists, or the slot is narrowed to that boolean.

What this change ships

Inventory (10 items) — 7 justified

Intent: let a user hand a chat to their Kiro Cloud Sessions sandbox from the environment chip. ADDITION.

  1. Settings has a Kiro Cloud Sessions section for saving cloud profiles — justified
  2. The chip can bind a chat to a cloud profile, which runs Kiro's own agent there — justified
  3. A policy capability that allows cloud starts on a governed host — justified
  4. Editing the default profile now shows a code-exec approval card — justified
  5. Saving one id for both a Coder and a Kiro Cloud profile is refused with a 400 — symptom-level: lookup ignores the provider the row records
  6. Shutdown and drain no longer cancel a running cloud turn — justified
  7. Approvals pending when the chat reopens are held and offered again — oversized
  8. When a reopen fails, an error row says a new sandbox started — one consumer, generalized
  9. The Files, Git and terminal views show a notice, and routes answer 409 cloud_sandbox — justified
  10. A cloud session nothing recorded is deleted; teardown otherwise keeps the cloud session — justified

[FIRST-PRINCIPLES-REVIEWED] 4c92545

Description read

[DESCRIPTION-READ] cff6c78e189dd70dfa22641c678133be0506ad48b97c3f46ee157ef0f9f59d82

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 commented Oct 16, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🟡 CONCERNS

UX-level review of 4c925456c6f85f4a6524d1ec28bd9c164add7b61 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

First-time users can't tell what to type in the Repositories field, and a stale-placement race flashes a red error for a chat that is fine.

Watch

  • Repositories field is unclear. The hint says "as owner/name (e.g. acme/web)" but never says which service the repositories live on. The blind reader rated the field "a guess", said they'd use it only "maybe", and wrote: "I don't know what a 'repository' is or where to find the 'owner/name'." This field is the main input for a cloud profile and every new user fills it in. The cost is friction, not failure. Fix: name the host in the hint.
  • Red error for a non-failure. When the recorded placement is stale, cloudSandboxRefused shows as an ErrorNotice before the info notice replaces it. Both say nearly the same thing. The reader on shot-02: "I can't tell whether something actually went wrong." This only happens in a race, but it makes the user doubt a chat that works. Fix: show the info notice while the placement is being re-read, not an error.
  • Replay notice reads as lost work. The reader on shot-03: "I'd think I had lost work… unclear whether 'session' and 'sandbox' are the same thing." It is rare but alarming. Fix: use one word for the thing, and make app.kiro.dev a link the user can click to reach the earlier session.

Suggestions

  • The existing Coder button "Add profile" sits next to the new "Add Kiro Cloud profile", and the reader couldn't tell which kind of profile each one creates. Rename the Coder one to "Add Coder profile".

[UX-REVIEWED] 4c92545

Description read

[DESCRIPTION-READ] ecd6b6b6b3dd93a482e93fe503283a41f2d0adb8c3996c1d2dea3f5d54ffe719

That sha256 names everything this verdict read: the PR description, plus the 11 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 16, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 4c925456c6f85f4a6524d1ec28bd9c164add7b61 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 4c92545

Verdict parsed from the review's SHA-scoped output markers for commit 4c925456c6f85f4a6524d1ec28bd9c164add7b61.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 4c925456c6f85f4a6524d1ec28bd9c164add7b61: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 16, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 4c925456c6f85f4a6524d1ec28bd9c164add7b61 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 4c92545

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 4c925456c6f85f4a6524d1ec28bd9c164add7b61: <one-sentence reason>

@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 16, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch from 6b91cb7 to 47b1810 Compare October 19, 2026 00:03
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-kiro-cloud branch from b66eeee to f184c66 Compare October 19, 2026 00:27
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch 2 times, most recently from 5da3123 to 9c027ca Compare October 19, 2026 00:35
@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 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch from 9c027ca to 91efcec Compare October 19, 2026 01:11
@kyleseaman
kyleseaman force-pushed the feat/session-environments-kiro-cloud branch from f184c66 to 42c29b1 Compare October 19, 2026 01:46
@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Oct 19, 2026
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed merge conflict Branch has merge conflicts with its base — author must resolve before merge labels Oct 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch from cde57d6 to cf8307a Compare October 19, 2026 06:50
@github-actions github-actions Bot added merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch from cf8307a to f87434b Compare October 19, 2026 07:23
@kyleseaman
kyleseaman force-pushed the feat/session-environments-kiro-cloud branch from 5d02699 to 25aeda8 Compare October 19, 2026 07:29
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: passed Eligible automated validation passed for the current revision merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Oct 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch from f87434b to 60d58f7 Compare October 19, 2026 07:57
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 19, 2026
@kyleseaman
kyleseaman force-pushed the feat/session-environments-git branch from 60d58f7 to c017d8c Compare October 19, 2026 08:20
A Kiro Cloud profile asks KAS to place the session in a Kiro cloud sandbox
with the _meta.kiro keys kiro-cli's --cloud flag sends, and refuses to start
rather than run locally when KAS does not offer cloud sandboxes.
@iamwhatever

Copy link
Copy Markdown
Contributor

HOLD at 4c92545. This needs a human reviewer, and it cannot be approved yet in any case.

What blocks approval:

  1. The base is feat/session-environments-git, not main. It clears once the stack below lands and this PR is retargeted to main and reviewed again at that head.
  2. It is a security-sensitive change, which our rules reserve for a human reviewer. It adds the capabilities.kiro_cloud_sessions governance row and a launch gate that lets a chat run outside Kiro Crew's tool gate (session_launch.py#L288, catalog.py#L405). It also adds paths that delete cloud sessions and kill runtimes. I found no definite defect in the gate: it fails closed on read errors and unreadable profiles, and it audits each decision.
  3. AI items on this head have no reply. The First Principles CONCERNS has three items (the replayed-approval hold, the profile-id collision code, startup_notice). The UX CONCERNS has three Watch items and one suggestion. The Design check asks for a real KAS confirmation of executionTarget: cloud-sandbox and sessionSource: remote. Each needs a fix or a disposition.

Checks are green on this head.

— crew-pr-reviewer (r113)

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • rebutted span=3059173b2655 (replayed-approval hold)

KAS replays a pending permission on session/load. Without the hold, _answer_unroutable_permission rejects it: the agent sees its tool call denied and the user never saw the approval card.
The owner stay-true requirement for this PR is that a replayed approval reaches the normal approval path. The hold waits CLOUD_PERMISSION_HOLD_SECS for the handle, then dispatches through it; on timeout it still takes the same audited reject. Nothing is auto-approved.
It is cloud-only (kas_cloud.py; the shared drain only reads a set no other harness fills, per H13). Dropping it is a Goal change, which this item leaves to the owner.

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • rebutted span=2bf030dde8a5 (profile id collision check)

A (provider, id) lookup cannot resolve every reader: session_environments.default_profile is a bare id (validate_default_profile), and the launch path, picker and choice lock resolve it with profile_provider(section, id) before any choice row exists.
With a shared id, that default would name two profiles and the start would have to pick one silently. Refusing the shared id at save time (ProfileIdCollision) is the one place that keeps every id unambiguous, per the owner ruling that profile ids are unique across providers.

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • rebutted span=8619e27395f1 (LLMProvider.startup_notice slot)

The tuple is (row kind, channel fallback text). A cloud_session_replayed boolean on LLMProvider would move that cloud kind and its English text into the shared reader _surface_startup_notice (dashboard/chat_turn/turn_spec_hooks.py), a cloud branch in shared dashboard code that the H13 harness-parity rule forbids and that GPT blocked twice on earlier heads.
As shipped, only the cloud start path (acp_runtime_start.py) knows the kind and text; the shared reader appends whatever row the provider recorded.

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • fixed span=c1b9dcc86979 (Repositories field unclear)

The hint now says where the repositories live: Repositories on the source provider (such as GitHub) connected to your Kiro account, comma-separated as owner/name (e.g. acme/web), at most {{max}}. Leave it empty for an empty workspace.
All 12 catalogs plus the regenerated en-XA. Commit bfe7a9d63a, local: it is published with the next restack push, which the stack conductor is holding until the base PR is green-candidate.

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • fixed span=fd2b3922bc01 (replay notice reads as lost work)

One word for the thing: the notice now says a new cloud session started with this conversation carried over, files and changes from the earlier session are not in it, and the earlier session is still listed in Kiro Web (app.kiro.dev). No more sandbox vs session.
All 12 catalogs, the backend channel fallback (acp_runtime_start.py) and both tests. Commit bfe7a9d63a, local, published with the held restack push.
The link part stays plain text: ErrorCard renders every error kind as text, and app.kiro.dev is spelled out in full.

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • rebutted span=c1c23dc1691d (red error before the info notice)

A browse refused with 409 cloud_sandbox is a rejected request, and the blocking errors-use-error-notice rule requires ErrorNotice for it; GPT blocked the neutral rendering on 4a7d733.
The refusal already triggers one re-read of the chat placement, and the info notice replaces the error as soon as that read reports a cloud chat. The window is only the stale-placement race the finding describes.

@kyleseaman

Copy link
Copy Markdown
Contributor Author
  • rebutted span=1da9957ed3c8 (rename Add profile to Add Coder profile)

The Add profile button, its catalog key and the tests that assert its label belong to the base PR #19833 (Coder), which is mid-restack. Renaming it from this stacked PR would edit p1 strings and tests and conflict on every p1 push.
Passed to the stack conductor for the #19833 owner to take there.

@iamwhatever

Copy link
Copy Markdown
Contributor

HOLD at 4c92545. This still needs a human reviewer and cannot be approved yet.

Thanks for the dispositions. The First Principles items and two UX items are rebutted with reasons. Two UX items are marked fixed in commit bfe7a9d63a, which is not on this head, so they count once that push lands.

What still blocks approval:

  1. The base is feat/session-environments-git, not main. It clears once the stack below lands and this PR is retargeted to main and reviewed again at that head.
  2. It is a security-sensitive change, which our rules reserve for a human. It adds the capabilities.kiro_cloud_sessions governance row and a launch gate that lets a chat run outside Kiro Crew's tool gate (session_launch.py#L288-L378), plus new permission-hold and runtime-kill paths. I found no definite defect in the gate: it fails closed and audits each decision.
  3. The Design Review asks for a live KAS check that executionTarget: cloud-sandbox starts in the sandbox and reopens with sessionSource: remote. It has no reply, and the description lists that live run as still to do. Please run it or reply.

Checks are green on this head.

— crew-pr-reviewer (r116)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants