Skip to content

docs(rfc): app SDK browser, a permission-gated view, read and annotate surface - #38322

Merged
iamwhatever merged 1 commit into
mainfrom
rfc/app-sdk-browser
Nov 18, 2026
Merged

iamwhatever merged 1 commit into
mainfrom
rfc/app-sdk-browser

Conversation

@chenmingwei23

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Let an App Kit app show, read and annotate the browser page that its Bot uses, through a public SDK surface instead of private imports and patches.

Crew has a browser (the browser tool, a live view, user takeover). An app has none of it: ctx has no browser and @kirocrew/app-sdk exports no browser view. The one app that needs it, Browser Collab, works only by importing about a dozen private gateway modules and by carrying three patches against Crew. This RFC turns that into a manifest permission, a backend ctx.browser, and a front-end BrowserView, behind one preview flag that is off by default.

Why it matters

Every app that wants to show or read a page has the same problem. A patch that one user carries cannot be reviewed, versioned or removed, and it gives one app a back door that others are denied. A public surface puts the identity checks in one place.

Not a goal

  • inject (writing content into a page). Phase 1 has none; the RFC lists the minimum a second RFC must meet.
  • Third-party app agent tools. The accepted rfc-app-agent-tools-third-party-grant.md and feat(apps): list a third-party app's agentTools on an opt-in prompted kirocrew-apps server #17015 own them. This RFC adds no server, grant family or flag for them.
  • Isolating an app from the gateway. That is rfc-app-sandbox-isolation.md.
  • Code. This PR is a design document only; the implementation PRs start after a maintainer signs it.

What changed (motivation -> approach -> change)

Adds docs/request-for-change/rfc-app-sdk-browser.md (status draft) and one index row in the RFC README.

Approach, chosen over the alternatives in the document:

  • One gateway function, run_browser_op, extracted from api_browser_op, that the route and ctx.browser both call. The URL check on the requested address, the lock, the takeover hold, the tab binding and the on/off check then exist once.
  • BrowserView wraps the existing live-view component and hooks. No new channel, no new route.
  • Slot ownership checks the key prefix and the _app stamp, and tabs come from LiveHub.targets_of.
  • The grant is a staged family in app-grants, approved by the owner after install. A verb added by an update waits for the owner.
  • The trust model follows chat_slot_sdk.py: a grant is consent to a stable interface, not isolation; execution trust is the boundary.
  • Third-party tools depend on feat(apps): list a third-party app's agentTools on an opt-in prompted kirocrew-apps server #17015 and the accepted RFC. Browser Collab companion agent is proposed as the acceptance case for that RFC phase 2. User Trust, YOLO and always-trust keep working; server config and the app itself can never waive approval.
  • Docs ride with each implementation PR. SDK level and RFC phase numbering are coordinated with the third-party grant RFC.

Backwards compatibility

Removes nothing: a docs-only change (one new file and one README row). The RFC itself proposes only additive changes (an optional manifest field, a flag that is off, an SDK level bump).

Tests

N/A - docs only. Every claim about main was checked against origin/main at 5d18d11. Proof: not provable -- docs or config only.

Manual verification

N/A - no behavior changes. The RFC text was read against the cited code paths on main.

Related Issues

Builds on #17015 and the accepted rfc-app-agent-tools-third-party-grant.md (tracking issue #32727).

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title
  • Existing tests pass and new tests added for new functionality (N/A, docs only)
  • 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 Apache-2.0 license.

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Nov 17, 2026
@github-actions

github-actions Bot commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ⏭️ skipped

Revision ee5e08bed4884ee7f3821b3af21486521fcefa3c touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

@github-actions

github-actions Bot commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ⏭️ skipped

Revision ee5e08bed4884ee7f3821b3af21486521fcefa3c ships no reviewable capability (docs, tests or generated files only), so there is nothing to inventory. Does not block merge.

@github-actions

github-actions Bot commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of ee5e08bed4884ee7f3821b3af21486521fcefa3c — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: this is a draft RFC that changes only docs and adds no code, surface or take-away.

Notes

  • §2 moves the requested-URL check into run_browser_op but keeps a copy in the shim. That leaves two copies, which the same section says it wants to avoid ("no gate has two copies"). Either describe the shim's copy as an early-fail convenience that the gateway check can override, or drop it.

[DESIGN-REVIEWED] ee5e08b

Description read

[DESCRIPTION-READ] c7ac46900bb3cd28b4a7f0c616120c24c5a187ba89929e6df1e5cabdf78a1ed7

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 Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of ee5e08bed4884ee7f3821b3af21486521fcefa3c and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] ee5e08b

False positive or not applicable? A repository writer can comment:
/ai-review override gpt ee5e08bed4884ee7f3821b3af21486521fcefa3c false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed ee5e08bed4884ee7f3821b3af21486521fcefa3c — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] ee5e08b

Verdict parsed from the review's SHA-scoped output markers for commit ee5e08bed4884ee7f3821b3af21486521fcefa3c.

False positive or not applicable? A repository writer can comment:
/ai-review override fable ee5e08bed4884ee7f3821b3af21486521fcefa3c false-positive: <reason, 20 to 500 characters>

@chenmingwei23
chenmingwei23 force-pushed the rfc/app-sdk-browser branch 2 times, most recently from eb28483 to 8f055a0 Compare November 17, 2026 09:04
@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Nov 17, 2026
@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

This adds a new concept, so it needs a design sign-off from a maintainer before it can merge. A maintainer will post the decision here.

To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/).

Review result on this head: nothing blocks. The diff changes docs only: the new draft rfc-app-sdk-browser.md and one index row. We checked the RFC's claims about main against the code, and they hold. OPS has 17 entries and no evaluate. OpContext.url_allowed allows every URL by default. api_browser_op, LiveHub.targets_of and VISIBLE_TABS_CAP exist where the RFC says they do. APP_SDK_LEVEL is 9 in both mirrors. The chat_slot_sdk.py trust wording matches. A strength of the RFC: it sees that an OpContext built by hand would skip the requested-URL check. Its answer is one run_browser_op door, with named tests that fail if the extraction is skipped.

Optional notes for the signer:

  • The act verb (navigate, click, type) goes past the stated goal of show, read and annotate. The RFC's own Open question 2 doubts that the first app needs it. You could defer it, as the RFC already defers inject.
  • Open feat(apps): let an app's own agent browse its approved sites unasked #38333 adds permissions.browserSites and also claims SDK level 10. The RFC does not mention it. One line that says how permissions.browser and permissions.browserSites relate would let the signer decide both shapes at once.
  • The per-slot tab cap in section 4.3 does not say what it limits. Does it bound only the SDK's view of tabs, or also how many tabs the Bot may open in an app slot? If it limits the Bot, list that under Backward compatibility.

@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: passed Eligible automated validation passed for the current revision labels Nov 17, 2026
@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Nov 18, 2026
@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Code change approver recommendation: A person should decide

Why: the decider estimates an 8% chance that this change introduces a defect, within the 12% limit for approving on its own, but it changes 550 lines, more than the 400 it may approve on its own.

Docs-only draft RFC plus one README index row. I checked its claims about main against the code: the api_browser_op gates, the allow-all OpContext.url_allowed default, the requested-URL check that runs only in the shim, APP_SDK_LEVEL=10, the chat_slot_sdk trust wording, AppHost having no iframe, the staged grant helpers, the browserSites support and the governance ceiling filter in bridges_agents. All hold. The trust model says plainly that the grant is consent and not isolation, and that execution trust is the boundary. It moves no boundary, because app pages already run with the owner's cookie. The only problems are wording, so I approve.

Tier: 2, held for a person: design-doc-gate: D.

An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit ee5e08bed4 · 550 changed lines. If this recommendation is wrong, reply here: replies are how the approver is evaluated.

@iamwhatever
iamwhatever merged commit 8fe7ca4 into main Nov 18, 2026
98 of 99 checks passed
@iamwhatever
iamwhatever deleted the rfc/app-sdk-browser branch November 18, 2026 05:48
@bolichen97

Copy link
Copy Markdown
Contributor

Post-merge input for the implementation PRs: this RFC merged at 21:48 PST with status still draft.

Review input on the RFC as merged (ee5e08bed4), checked against main 45dc4dd2b5. This is design input for the implementation PRs, not a sign-off.

The claims about main hold: OPS is a wire enum with no evaluate (browser_harness/ops.py:56-81); OpContext.url_allowed defaults to allow-all (ops.py:218) and the gateway judges only the committed URL (ops.py:687); the shim checks the requested URL (mcp_tools/browser.py:499); APP_SDK_LEVEL = 10 in both mirrors; LiveHub.targets_of and VISIBLE_TABS_CAP = 32 are at live_hub.py:871 and :97; bridges_agents._ceiling_filtered_allowed filters only against the ceiling (apps/bridges_agents.py:202). Four points the implementation should settle, then the open questions.

A. #38333 already put a requested-URL check in the gateway, for app slots. api_browser_op now reads session_sites and refuses an off-site navigate / tabs new before the op runs (dashboard/handlers/browser_ops.py:637-651), and wires _site_gate(sites) into url_allowed (:293). So the §2 table row ("only in the shim") is true only for sessions with no browserSites. Since ctx.browser will run through the same body, please state what an app slot's navigate is held to: its approved browserSites when it has them, and the plain url_gate otherwise. That second case is the one that matters (see Q2).

B. run_browser_op(session_key, op, args) cannot take the body as is. The body reads request in several places: internal_auth (browser_ops.py:605), session_app reads request.app["state"]._slots (:376-381), and _audit, _held and _start_denied all take the request (:125, :777, :472). The RFC should say what replaces request (the dashboard state and an audit sink, passed in), and that internal_auth stays on the route only. Otherwise step 1b will either thread a fake request through, or drop the site lookup for SDK calls without anyone noticing.

C. Register browser the way browserSites is, not the way api / events are. api and events are approved at install. The precedent for "install approves nothing" is browserSites: register_permission_grants(..., explicit_only=True) (apps/point_grants.py:224) plus a row in _OWNER_APPROVED_FAMILIES (apps/manager_records.py:1145). Without explicit_only, an old record with no browser key approves what the earlier manifest declared (_approved_grants_after_manifest_change, manager_records.py:1158-1191). §1 should name both.

D. The shim's URL copy (Design Review note): agreed. Describe it as an early-fail convenience, so the gateway check stays the one that decides.

Open questions

  1. Per-slot tab cap of 8. There is no per-session tab cap on main today. If it lives in run_browser_op, it also caps the Bot's own tabs new in every slot, because the Bot and the SDK share that door. PagePort has no tabs new, so the SDK only gets tabs through popups (F3). My view: leave it out of phase 1. If it is wanted, make it a Bot-wide step-0 change with its own PR and a Backward compatibility line. 8 is a fine number if it ships.
  2. act in phase 1. Defer it. The per-call card comes from the tool gate (hooks._tier_app_browser_sites and the PreToolUse gate, hooks.py:1505), not from api_browser_op. A backend ctx.browser.navigate/click/type skips the gate completely, so an approved act is a standing, card-free grant for the app's Python to drive the page. For an app without browserSites, that reaches any public URL (point A), which is wider than what feat(apps): let an app's own agent browse its approved sites unasked #38333 lets the app's own agent do. That conflicts with this RFC's own §7 line that an app cannot make its tools skip approval. If act comes back later, I'd require approved browserSites for it and hold navigate to them, so one owner-approved site list covers both paths.
  3. text and pick from snapshot. pick is not enough: snapshot rows carry no coordinates, and the walker uses elementFromPoint only to test whether an element is covered (snapshot_walker.js:203). Decide now instead of in step 1b. Plan one read-only walker method (point to ref, ref to text) behind a new OPS entry. Note that OPS is the Bot tool's wire enum (ops.py:56-60), so a new entry is also a new Bot verb. Either say that is intended, or keep SDK-only ops out of the advertised enum.
  4. Where the "never" clarification goes. Put it in rfc-app-agent-tools-third-party-grant.md (the may_skip_gate sentence, line 239). Readers of that RFC will not open this one.
  5. Who builds the grant RFC phases. That is for the feat(apps): list a third-party app's agentTools on an opt-in prompted kirocrew-apps server #17015 authors (still open). One suggestion: do not make this RFC's step 2 acceptance wait on companion tools. Accept view, read and annotate on their own, and move the companion-agent check to step 3.

One process note: the file merged with status: draft and an empty doc-pr:. If this merge was meant as the design sign-off, a one-line follow-up that flips the status to accepted would make that clear in the record. If it was not, the status is right as it is.

A team member must sign off on this design; this comment is review input only.

AI-assisted, sent by Bolin Review bot

chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.
iamwhatever pushed a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
chenmingwei23 pushed a commit that referenced this pull request Nov 18, 2026
chenmingwei23 added a commit that referenced this pull request Nov 18, 2026
…n app's own slot (#39020)

Implements step 1c of RFC #38322. BrowserView wraps the dashboard's live
view (BrowserLiveCardCanvas, browserLiveStore) and adds no channel or route.
Behind the app_browser_sdk preview; a slot the gateway did not stamp with the
app's name shows denied. A hand-back note no agent op took is sent once:
by a mounted chat surface (which now calls preventDefault), else by the
store. APP_SDK_LEVEL 10 -> 11.

Co-authored-by: Raymond Chen <raymondchen625@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

readiness: passed Eligible automated validation passed for the current revision

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants