Skip to content

fix(cloud): report a found but refused AWS CLI in the pre-flight instead of not installed - #22162

Merged
chenmingwei23 merged 1 commit into
mainfrom
fix/20238-aws-cli-declined
Oct 20, 2026
Merged

chenmingwei23 merged 1 commit into
mainfrom
fix/20238-aws-cli-declined

Conversation

@iamwhatever

@iamwhatever iamwhatever commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: When the gateway finds an aws CLI but its safety check refuses to run it, the Remote Crew AWS pre-flight says where it was found and why, the way the Session Manager plugin row does, instead of reporting the CLI as missing.

The plugin row tells "not installed" apart from "installed but refused" (#19960). The AWS CLI does not. A refused aws reads as missing: the dashboard's credentials row shows aws CLI not found — install it first, and kirocrew cloud doctor prints aws CLI not found — install it. Both send the user to reinstall the same refused file.

Why it matters

The common case is a Homebrew install owned by a second account (Workbrew), which the check refuses on purpose. Reinstalling the CLI the same way cannot fix that, so the user follows the hint, re-checks, and gets the same row with no clue why.

Not a goal

  • Changing which binaries the safety check accepts, or which directories the AWS child's PATH holds.
  • Offering a copyable fix command for the CLI. None is known to clear its refusal; the plugin's chown only applies inside AWS's plugin install tree.
  • Rewording the existing plugin_refused_* sentences or the plugin row.

What changed (motivation → approach → change)

A refused CLI looked missing because only the plugin read the resolver's refusal list. The resolver already records every refused candidate (resolve_aws_tool_bin_with_refusals); nothing asked it about aws.

aws_cli_refusal() in cloud/ssm.py reads that list for aws, sharing one body with session_manager_plugin_refusal() and its refusal codes. It never returns a command.

Surface Before After
Dashboard credentials row (RemoteCrewPanel) aws CLI not found — install it first … "Your AWS credentials were not checked." + the plugin row's own sentence ("Found at , but Kiro Crew will not run it: …") for the reason (path, owner, root, strict, world-writable), the check's raw reason under it, the plugin row's next step for that reason, the guide link. No command.
/api/cloud/preflight (dashboard/handlers_cloud.py, typed in api/client/cloud.ts) — aws_cli_refused_path/_reason/_code/_owner/_owner_uid, read only when credentials did not resolve
kirocrew cloud doctor (cli_cloud.py) aws CLI not found — install it (…) aws CLI found at <path> but refused: <reason> + a next step for that reason; the "AWS not reachable" line says the profile was not checked instead of profile/helper advice

remoteCrewPluginRefusal.ts turns its plugin-only helpers into one shared path for both binaries; the plugin row's output is unchanged. One new string, aws_cli_refused, is in every locale and in en.context.json.

The fix sits at the two pre-flight surfaces (dashboard credentials row, kirocrew cloud doctor). run_aws's shared "aws CLI not found" 127 message, which 35 call sites share, is left in place. The refusal fields are filled only when credentials did not resolve, and the dashboard shows them only then, so a refusal never sits beside a working profile. A next step for a file owned by another account, shared by both rows, is #22203. The Remote Crew on EC2 guide's pre-flight entry gains a bullet for the refused CLI.

Backwards compatibility

Compatible: the preflight response only gains optional fields, the plugin row renders the same copy, and a CLI that is found, accepted or truly absent shows what it showed before.

Removes nothing: no key, field, function or path is removed or renamed; pluginRefusalSentence, pluginRefusalCopy and pluginRefusalReport keep their signatures.

Tests

  • test/test_cloud_ssm.py::TestAwsCliRefusal: absent CLI → none; accepted → none; the real check's third-account refusal → path, other_owner, uid and no command, even inside the plugin's chown tree; real root-gateway refusal → root_gateway.
  • test/test_cloud_handlers.py: the preflight carries the five CLI fields when credentials fail, and never reads the refusal once they resolve.
  • test/test_cloud_cli.py: the doctor reports found-but-refused with the next step for each of other_owner, world_writable, root_gateway, strict_policy, with no install link and no profile advice.
  • website/src/test/RemoteCrewPanelPluginRefusal.test.tsx: the rendered credentials row leads with the refusal, keeps the raw reason, has no not-found text, no command, gives the plugin row's ownership step, and links the guide; world-writable keeps the ownership step; the root step; the agent report; null once credentials resolve.

All 15 new tests fail on origin/main (the backend ones with has no attribute 'aws_cli_refusal', the frontend ones on the missing export).

Determinism: the 10 backend tests ran green 10 times in a row and once in a shuffled order (seed 20238). The frontend file ran green 5 times and once with --sequence.shuffle --sequence.seed=20238.

Manual verification

Captured the built SPA with /api/** answered from fixtures (a local Playwright harness, not committed), asserting the row text before each capture. The gateway's probe of a real refused install is covered by the backend tests that run the real safety check.

Screenshots / video

Before (same situation on main): the row says the CLI is not found.

Before: aws CLI not found

After, a CLI owned by another account (dark):

After: refused, other owner

After, a root gateway (light):

After: refused, root gateway

Related Issues

Closes #20238

Pattern harvest

Not generalizable: the resolver's refusal list was read for one of the two binaries it serves; this PR reads it for the other, through the same helper.

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

@iamwhatever
iamwhatever requested review from a team and KiroCrewCaptain as code owners October 19, 2026 12:52
@iamwhatever
iamwhatever requested a review from buluoray October 19, 2026 12:52
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 19, 2026
@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 5afb36960977de936ead512089b9c6426874a9a0 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Check by hand that the reachability probe and aws_cli_refusal() run separate resolver walks, so an install fixed between them gets a stale row that a re-check clears.

[DESIGN-REVIEWED] 5afb369

Description read

[DESCRIPTION-READ] 10048aa43f496bc12d683f95f64c0e2038777c52394ff666f4df0cf0a39e4a29

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 19, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 5afb36960977de936ead512089b9c6426874a9a0 — 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: PASS

Check before merge: run_aws's shared "aws CLI not found" message still misreports a refused CLI everywhere except these two pre-flight surfaces. The description defers that on purpose.

What this change ships

Inventory (8 items) — 8 justified

Intent: when the aws CLI is found but refused, the Remote Crew pre-flight says where it was found and why it was refused, instead of sending the user to reinstall the same file (#20238). This is a FIX.

  1. The dashboard credentials row says the credentials were not checked, gives the found-but-refused diagnosis and the next step, and drops the "not found" install link — justified
  2. kirocrew cloud doctor prints "aws CLI found at but refused: " plus a next step for that refusal code — justified
  3. When the CLI was refused, the doctor's "AWS not reachable" line drops the profile and helper advice — justified
  4. /api/cloud/preflight gains five optional aws_cli_refused_* fields, filled only when credentials did not resolve — justified
  5. "Ask the agent" on the refused-CLI notice sends its own report id, remote-crew-aws-cli-refused — justified
  6. A new aws_cli_refused string in every locale — justified
  7. The plugin-only frontend helpers become one shared path for both binaries, and the plugin row's output stays the same — justified
  8. The EC2 guide's pre-flight entry gains a bullet for a refused CLI — justified

[FIRST-PRINCIPLES-REVIEWED] 5afb369

Description read

[DESCRIPTION-READ] 5bb2aefcf02f7c42e512575c964c0152b01bfa1b60448ab8906bb28fcd78e448

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 19, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🟡 CONCERNS

UX-level review of 5afb36960977de936ead512089b9c6426874a9a0 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

The refused-CLI row names a fix ("fix the file's ownership or permissions") that the blind reader could not act on. The raw English reason repeats the sentence above it.

Watch

  • The next step doesn't say what to do. For other_owner the row shows the plugin's shared step, "Fix the file's ownership or permissions, then re-check." The blind reader said they had "no idea how to 'fix ownership'", and on the root case, "I don't know how to do it". "Permissions" doesn't apply to an ownership refusal at all. So the user's only real paths are the agent and an EC2 guide. That guide's name made the reader doubt it applies to them: "I don't know what 'Remote Crew' or 'EC2' are, or whether they apply to me". Severity: rare state, but hitting it stalls the task, and Remote Crew AWS pre-flight: give a refused plugin or CLI owned by another account a next step that clears it #22203 is deferred. Smallest fix: link the guide's anchor as "How to fix a refused AWS CLI" instead of the guide's title.
  • The monospace detail repeats the sentence above it. The detail line ("executable is owned by another user (uid 502)", "provider execution is disabled for a root gateway") says the same thing again in harder words. The reader called it "the same message again in computer language" and couldn't "tell what the second line is for." Severity: shows every time the row appears, adds friction but doesn't block. Smallest fix: drop failureDetail for codes that have a translated sentence. The agent report already carries the raw reason.
  • The advice points at a control above it. "If you are not sure how, use 'Ask the agent'" refers to a link that sits above the advice and is styled red like the error. The reader saw it "as part of the error, not a helpful button." Severity: low, cosmetic, but on the main path. Smallest fix: place the steps before the agent link (component-level, shared with the plugin row).

[UX-REVIEWED] 5afb369

Description read

[DESCRIPTION-READ] 9b6c2b81ac492e04abf88d20ebd7e896ad40dcfe00e58c74107b729cb4d62965

That sha256 names everything this verdict read: the PR description, plus the 6 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 19, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 5afb36960977de936ead512089b9c6426874a9a0 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 5afb369

Verdict parsed from the review's SHA-scoped output markers for commit 5afb36960977de936ead512089b9c6426874a9a0.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 5afb36960977de936ead512089b9c6426874a9a0: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 5afb36960977de936ead512089b9c6426874a9a0 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 5afb369

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 5afb36960977de936ead512089b9c6426874a9a0: <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 19, 2026
@iamwhatever
iamwhatever force-pushed the fix/20238-aws-cli-declined branch from b6f63a6 to deb8cfd Compare October 19, 2026 13:28
@iamwhatever

iamwhatever commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor Author

self-added: no
mechanism: CLI-specific next step for other_owner

  • span=35e6c86ef1ce The next step for other_owner sends Workbrew users to the wrong fix -- fixed in deb8cfd

fixed: for an aws owned by another account the row and kirocrew cloud doctor now say "Install the AWS CLI with AWS's own installer, then re-check."
Every other refusal code keeps its step; the plugin row is unchanged.
The resolver keeps walking install dirs after a refusal, so an AWS-installed copy in a later dir is picked up on re-check.

@iamwhatever

iamwhatever commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor Author

self-added: no

  • span=51713a349daa Add a CLI-specific other_owner step in the panel and the doctor -- fixed in deb8cfd

fixed: new key aws_cli_refused_next_other_owner in all 13 locales and en.context.json; doctor _REFUSAL_NEXT_STEP['other_owner'].
The wording drops "it is owned by root": AWS's macOS plugin package is known not to install root-owned, and the CLI package was not verified, so the step claims only what it does.

@iamwhatever

iamwhatever commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor Author

self-added: no

  • span=acd862772644 Reword aws_cli_refused to say who refused -- rebutted

rebutted: the sentence that follows the lead, in the same alert, is "Found at , but Kiro Crew will not run it: ", so who refused and why are on screen.
The suggested lead would say "Kiro Crew won't run" twice in a row. The lead's job is the consequence: credentials were not checked.

@iamwhatever
iamwhatever force-pushed the fix/20238-aws-cli-declined branch from deb8cfd to 131485a Compare October 19, 2026 14:30
@iamwhatever

Copy link
Copy Markdown
Contributor Author

self-added: yes
mechanism: CLI-specific next step for other_owner

  • span=ff25586d06af The other-account step reads as "reinstall what you already have" -- fixed in 131485a

fixed: the step now reads "Install a separate copy with AWS's own installer, then re-check. If the new copy is refused too, use “Ask the agent”." in all 13 locales.
The doctor says the same; the guide says the copy lands in a separate install folder that the check walks after the refused one.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=0a6be0e7eff1 "Refused" doesn't say who refused -- rebutted (same ruling as span acd862772644 on b6f63a6)

rebutted: the next sentence in the same alert is "Found at , but Kiro Crew will not run it: ", so who and why are on screen.
The lead states the consequence (credentials not checked). Leaving the wording to the maintainer at review.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=08e5741596fa Lead aws_cli_refused with the outcome and who blocked it -- rebutted (same ruling as span acd862772644 on b6f63a6)

rebutted: the suggested lead "Kiro Crew won't run this AWS CLI" is followed at once by "but Kiro Crew will not run it", so the row would say it twice.
No copy change; the maintainer can override the wording at review.

@iamwhatever

Copy link
Copy Markdown
Contributor Author

self-added: yes

  • span=bcbcaec582e9 other_owner also covers an executable parent owned by another account, where a fresh AWS install can be refused again -- fixed in 131485a

fixed: the step no longer promises a fix. Dashboard: "If the new copy is refused too, use “Ask the agent”." Doctor: "If a copy is still refused, this line names it."
The guide names the case: the folder the copy lands in (e.g. /usr/local/bin under an Intel Homebrew run by a second account) can belong to the same account.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • The description says one new string, the diff adds two -- fixed (PR body)

fixed: the description now names both keys, aws_cli_refused and aws_cli_refused_next_other_owner.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • Check RemoteCrewPanel when both the CLI and the plugin are refused -- rebutted (verified, no change needed)

The credentials row reads only the aws_cli_refused_* fields through awsCliRefusalCopy; the plugin row reads only session_manager_plugin_refused_* through pluginRefusalCopy. The server fills both independently (api_cloud_preflight).
So each row names its own binary, path and next step; neither copy reads the other's fields.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 19, 2026
@iamwhatever

Copy link
Copy Markdown
Contributor Author

Main is red, not this PR: Harness Parity Gate fails on main's own Fast Gate runs (latest e5e6f1c, run 39274206222) with ACP_BACKENDS_MEMBER_CAPABILITIES [claude]: allowlisted but now proven; delete the entry. Await Fast Gate and Coverage Gate fail here only because of that. Fixes in flight: #22178 and #22186. This PR waits for one of them to merge, then rebases.

@iamwhatever

Copy link
Copy Markdown
Contributor Author

self-added: yes
mechanism: aws_cli_refused lead sentence

  • span=3cf41f5492ac "refused" points at the wrong party -- fixed in 5b2b7ed

fixed: the lead now reads "Kiro Crew did not run the AWS CLI it found, so your credentials were not checked." in all 13 locales ({{productName}} in the catalog).
The product, not AWS, is the subject, so the row no longer reads as AWS turning the user down.

@iamwhatever

Copy link
Copy Markdown
Contributor Author

deferred: the sentence comes from the shared plugin_refused_other_owner diagnosis, which the Session Manager plugin row shows too; rewording it here would change that row outside this goal.
The same sentence names the user ID and "the system administrator account (root)", which marks a computer account. #22203 rewords the other-owner case for both rows together.
The CLI-only "separate copy" step this item also named is removed in 5b2b7ed; the CLI row now uses the plugin row's next steps.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=da2ff4b677ea Lead with who did not run it -- fixed in 5b2b7ed

fixed: the lead is now the suggested shape, with the product as the subject: "Kiro Crew did not run the AWS CLI it found, so your credentials were not checked."

@iamwhatever

Copy link
Copy Markdown
Contributor Author

self-added: yes
mechanism: CLI-specific next step for other_owner (removed)

  • span=8956c6f6e544 Name the action and link the installer -- fixed in 5b2b7ed by removal

fixed: the round-3 retrospective removed the CLI-only aws_cli_refused_next_other_owner step; the string this item targets is gone.
The CLI row now shows the plugin row's next step for each reason plus the guide link. A step that clears the other-owner case for both rows is #22203.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=5f766242080e run_aws's shared "aws CLI not found" 127 message stays in place -- rebutted (scope), description updated

The fix sits at the two pre-flight surfaces (dashboard credentials row, kirocrew cloud doctor); the PR description now says so and that run_aws's 127 message, shared by 35 call sites, is left in place.
Changing run_aws would change every cloud command's error, which is new scope beyond the goal.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • Do the dashboard (reachable && account) gate and the handler's reachable gate agree? -- rebutted (verified, no change)

The handler fills aws_cli_refused_* only when reachable is false; the dashboard shows them only when reachable && account is false. Every response that carries a refusal is one the dashboard treats as failed.
A reachable response with no account carries no refusal fields, so it shows the generic failure; a refusal can never sit beside a working profile.

@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
@iamwhatever
iamwhatever force-pushed the fix/20238-aws-cli-declined branch from 5b2b7ed to 5afb369 Compare October 19, 2026 18:21
@iamwhatever

Copy link
Copy Markdown
Contributor Author

self-added: yes
mechanism: aws_cli_refused lead sentence

  • span=bf457af55494 The new lead and the reused sentence say the same thing -- fixed in 5afb369

fixed: the lead keeps only what the reused sentence lacks: "Your AWS credentials were not checked." followed by "Found at , but Kiro Crew will not run it: ." in all 13 locales.
"Did not run" now appears once, in the sentence that also says who and why.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=90b2a8ef7eb7 The heading does not match the body -- rebutted

The row is the existing "AWS credentials" check: its title and place in the list are unchanged, and a refused CLI is why that check could not run.
The lead now opens with exactly that link to the title ("Your AWS credentials were not checked."); moving the CLI into its own row is new UI beyond this goal.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=0ff3a1d4ac53 Keep only "Your AWS credentials were not checked." -- fixed in 5afb369

fixed: applied as suggested, in all 13 locales, the guide, the context note and the tests.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • Check one real Workbrew-owned aws on macOS before merge -- needs-a-decision (maintainer)

No macOS host with a Workbrew-managed Homebrew is available to this worker. The refused-CLI rows are proven by fixtures plus tests that run the real provenance check on a third-account-owned file and a root gateway.
Maintainer: please confirm on a real Workbrew Mac, or accept the fixture and real-check evidence.

@iamwhatever

Copy link
Copy Markdown
Contributor Author

The CLI row reuses the Session Manager plugin row's next step for each refusal reason, which is this PR's goal ("the way the plugin row does"). A CLI-only step was tried in rounds 0-2 and removed by the round-3 retrospective.
#22203 replaces the other-owner step for both rows at once, naming a fix known to clear it; changing it here would make the two rows disagree.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=4eec21a69ac1 The monospace detail repeats the sentence above it -- rebutted (shared plugin-row design)

The raw reason under the translated sentence is the plugin row's layout from #19960: the sentence is in the reader's language, the line under it is the safety check's own English words, and "Ask the agent" carries both.
This PR reuses that layout unchanged; dropping the line for one row only would split the two rows. The maintainer can rule on the shared layout.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=323bc658ed2e The advice points at a control above it -- rebutted (shared plugin-row design)

"Ask the agent" sits inside the error notice and the next step under it, the same order the plugin row has used since #19960; this PR adds no new control and moves none.
Severity is stated as low and cosmetic; a layout change belongs to both rows together, not to this fix.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • run_aws's shared "aws CLI not found" message still misreports a refused CLI outside the pre-flight -- rebutted (scope, stated in the description)

The description says the fix sits at the two pre-flight surfaces and leaves run_aws's 127 message, shared by 35 call sites, in place. Changing it changes every cloud command's error, beyond this goal.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • The reachability probe and aws_cli_refusal() run separate resolver walks -- rebutted (verified)

Yes: api_cloud_preflight calls aws_cli_refusal() after reachability_check, each walking the install dirs. If the install is fixed between the two, the row can show a refusal for a CLI that now passes; Re-check runs both walks again and clears it.
The window is one request long and needs the user to change file ownership mid-check; no state is kept between checks.

@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 Oct 19, 2026
@iamwhatever

Copy link
Copy Markdown
Contributor Author

Review-ready at 5afb369: PR Readiness passed, all 103 checks green, GPT and Opus clean, and every Design / First Principles / UX item on each head has its own reply.

Left for a maintainer:

  1. Merge and approval are yours; this worker does not merge, approve or arm auto-merge.
  2. First Principles asked for a check on a real Workbrew-owned aws on macOS. No such host is available here; the evidence is fixtures plus tests that run the real provenance check.
  3. UX's remaining points are about copy and layout the CLI row shares with the Session Manager plugin row (the next-step wording, the raw reason line, where "Ask the agent" sits). The shared other-owner step is Remote Crew AWS pre-flight: give a refused plugin or CLI owned by another account a next step that clears it #22203; the layout is your call.
  4. main has gained 2 commits since this head went green, in src/kiro_crew/gateway_lock.py, which this PR does not touch. I have not rebased, so the review lanes are not restarted; I can rebase on request.

@chenmingwei23
chenmingwei23 merged commit beb383f into main Oct 20, 2026
104 checks passed
@chenmingwei23
chenmingwei23 deleted the fix/20238-aws-cli-declined branch October 20, 2026 00:35
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 20, 2026
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.

Remote Crew AWS pre-flight: report a found but declined AWS CLI instead of not installed

2 participants