Repository navigation
fix(privacy): name store apps in the privacy chapter and bump the notice to 4 (#38145) - #38347
Conversation
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks: the version bump follows the rule both version files state ("bump both when the notice says something new"), and the new line covers What this change shipsInventory (6 items) — 5 justifiedIntent: tell every user, including anyone who already acknowledged notice v3, that store app names can now leave the machine. This is a FIX.
[FIRST-PRINCIPLES-REVIEWED] c0de3ad Description read[DESCRIPTION-READ] bd8524b0923d8058a4ecb26f4f8e23d46d08888c176a0c6839d5e78612caeb7b That sha256 names the PR description this verdict read. Recompute it with |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS No design-level concerns. This is one new line of copy plus the version bump the notice rule requires, and keeping Notes
[DESIGN-REVIEWED] c0de3ad Description read[DESCRIPTION-READ] 944a9175096d5a0c0f17523ff23e782bafcb638cbb037ba528e47a2956c46cdf 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 |
UX Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)UX-level review of UX-Verdict: PASS cannot evaluate: missing a screenshot of the new "What is collected" bullet. The PR description's image was not admitted, so no blind read ran. Evidence gaps
Notes
[UX-REVIEWED] c0de3ad Description read[DESCRIPTION-READ] 160ca68caf5ca1d05b1212d72cf1a4b6e1a14a99f340f05c9cea285bd3bdbefc That sha256 names everything this verdict read: the PR description, plus the 3 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 |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Thanks for adding this before the release that starts sending app names. One line in the new copy describes the wrong trigger, and it needs a fix before merge.
website/src/i18n/locales/en.manual.json:3990(collectedApps), plus the same text in the 12 other locales and the regex atwebsite/src/components/PrivacyChapter.test.tsx:80. The bullet says names are sent for apps "you open and use". On main, the only metric that carries a store app's public name iskirocrew.app.installed.record_app_installed(metrics/feature_usage.py) labels it withapp_public_id. It is called only fromstore_origin.count_first_install, andregistry_pipeline/install.pyruns that on a first registry install withviaset tostoreordefault. The default-apps pass (apps/default_installs.py) installs apps disabled withvia="default". Trigger: the default-apps pass installs a store app, or a user installs a store app and never opens it. Outcome: the app's name is still sent, while the chapter the user acknowledged says only apps they open and use are named.kirocrew.app.task.completedusesapp_label, which names only shipped built-ins, so no open or use event sends a store name today. Smallest fix: reword the key in all 13 locales so install is the trigger. For example: "Which built-in and store apps are installed or used, by their public store name. Apps you add from anywhere else are counted without a name." Then update the test regex to match. Can you make that change in this PR?
Optional:
- Notice v3 already shipped in v0.13.0-rc.2. Anyone who acknowledged v3 there will not see the chapter again when this line arrives. A maintainer may want to decide whether this counts as "something new" under the bump rule in
privacyNotice.ts. If it does, bump the version, or note in the body that those users are knowingly not shown the chapter again.
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Thanks for the quick rewording. The new line now names install as the trigger, it is in all 13 locales, and the test pins the English sentence. One item still blocks merge.
Earlier review on f78c939:
- Item 1 (the "open and use" trigger): fixed on this head. All 13 locales and the test regex now use the install wording.
- Optional item (notice version): the body now records the choice. The open question about it is below.
website/src/components/PrivacyChapter.tsx:22withwebsite/src/components/privacyNotice.ts:19andsrc/kiro_crew/beacon.py:154. This PR adds a new item to the "What is collected" list but leavesPRIVACY_NOTICE_VERSIONat 3. Both files say to bump it when the notice says something new. Trigger: a user on v0.13.0-rc.2, the tag that first carries version 3, pressed Keep on, so their storedprivacy_ack_versionis 3. They then upgrade to a build with #38157 and this PR. Outcome:firstRun.tsx:99shows the chapter again only whenprivacyAckVersion < PRIVACY_NOTICE_VERSION, and both values are 3. That user never sees the new bullet, butrecord_app_installednow sends store app names from their machine. Stable v0.12.2 upgraders and new installs are not affected. The body calls keeping version 3 an "owner decision", but no maintainer comment on this PR or on #38145 records that decision. The AI review BLOCK on this head also has no disposition and no override, so PR Readiness stays red. Smallest fix, pick one:- (a) Bump
PRIVACY_NOTICE_VERSIONto 4 inbeacon.pyandprivacyNotice.ts, add a line to thebeacon.pycomment saying what version 4 adds, and update the "both 3" text inmetrics.md.kiro_export.pygates on its ownNOTICE_VERSION = 3, so leave that at 3. Please check that Kiro telemetry export does not pause for v3 ackers, and add a test for it. The bump also shows the chapter and the headless notice once more to people who acknowledged v3. - (b) A maintainer records the decision on this PR to knowingly not show the chapter again to people who acknowledged v3 on the prerelease, and a repository writer posts an AI review override for that finding citing it. Then point the body at that comment.
- (a) Bump
Optional:
- The copy says store apps are named when "installed or used", but only an install sends a store app's name. A use event labels a store app as
private. I suggested that wording last round. It errs toward telling users more than is sent, so it is safe. If you want it exact, use something like "built-in apps installed or used, and store apps installed, by their public name" in all 13 locales and the test. - The change edits shipped UI copy, so
fix(privacy):orfeat(privacy):would describe it better thandocs:. The title gate accepts either.
…ice to 4 (#38145) kirocrew.app.installed is now a product metric (#38157), so a store app's public name can leave the machine on install. Add a 'What is collected' line that says so in all 13 locales and render it in PrivacyChapter: built-in apps are named when installed or used, store apps when installed, and apps added from anywhere else are counted without a name. Bump PRIVACY_NOTICE_VERSION 3 -> 4 in beacon.py and privacyNotice.ts, since the notice says something new and v3 already shipped in v0.13.0-rc.2. Keep kiro_export.NOTICE_VERSION at 3, so a v3 acker keeps exporting to Kiro telemetry, and pin that with a test. Update the version pins and the docs.
b3113d2 to
c0de3ad
Compare
|
Addressed on c0de3ad (one commit), for the review on b3113d2:
CI on c0de3ad: 87 pass, 0 fail, 0 pending; Opus 5.5 Review passes; PR Readiness success. |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Thanks for taking option (a). Nothing still blocks on this head.
Earlier review on b3113d2:
- Item 1 (notice version): fixed.
PRIVACY_NOTICE_VERSIONis 4 in bothbeacon.pyandprivacyNotice.ts.kiro_export.NOTICE_VERSIONstays 3.test_a_v3_ack_still_exports_after_the_notice_moves_onpins that people who acknowledged v3 keep exporting. - Optional copy wording: done in all 13 locales and the test regex. Store apps are now named only on install.
- Optional title: done.
What this review checked on c0de3ad: every reader of the version moved together, and no literal 3 that should have moved was left behind. The new bullet matches what main sends (app_public_id on install, app_label for built-ins only). The new tests are isolated and use no clock or sleep. PR Readiness passed. A good choice here is keeping the export gate apart from the notice version and adding a test that fails if anyone later moves the two together.
Optional:
docs/system-specs/modules/metrics.md:268still says the off offer applies while "notice version 3" is not yet acknowledged.docs/guides/telemetry-otlp-export.md:219still says "Keep off is not asked again". After the bump, a v3 Keep off user is asked once more. Both lines could name the current notice version instead.- Headless starts (
gateway --no-dashboard,run,chat) print the existing one-line notice, and the off offer line where it applies, once more for v3 users. Neither line names apps. A sentence under Backwards compatibility would record this. Right now the sibling sweep callsnotice.py"not affected". - The UX and Design lanes ran before the screenshot reached the body, so a lane re-run would let them see it. A narrow-width view (about 390px) would also show how the longer bullet wraps.
Problem / Motivation
Goal: The Privacy chapter tells every user which app names can leave the machine, and users who already acknowledged the previous notice see it once more.
kirocrew.app.installedis now inPRODUCT_METRICSon main (#38157,src/kiro_crew/metrics/product_metrics.py), so the public store name of a store app can leave the machine in the next release. The chapter's "What is collected" list does not say so, and notice v3 already shipped in v0.13.0-rc.2, so users who acknowledged v3 would never see a new line.Why it matters
The chapter is where users are told what leaves their machine. Without the line and a version bump, every user who acknowledged v3 is never told that app names are sent.
Not a goal
kiro_export.NOTICE_VERSIONstays 3).What changed (motivation -> approach -> change)
components.privacyChapter.collectedAppsinwebsite/src/i18n/locales/en.manual.json: "Which built-in apps are installed or used, and which store apps are installed, by their public store name. Apps you add from anywhere else are counted without a name." The wording follows what main sends: the only metric that carries a store app's public name iskirocrew.app.installed(record_app_installed, called fromstore_origin.count_first_install), which also fires for apps the default-apps pass installs;kirocrew.app.task.completednames only built-ins and sendsprivatefor store apps.npm run i18n:pseudo), the same file set recent keys use (fix(dashboard): show one calm offline state when a relay Crew's host or the local gateway stops #38103).website/src/components/PrivacyChapter.tsx: the key is added toCOLLECTED_KEYSaftercollectedFeatures.src/kiro_crew/beacon.pyandwebsite/src/components/privacyNotice.ts, per the rule both files state ("bump this whenever the notice says something new").beacon.pynotes what version 4 adds. A user whose storedprivacy_ack_versionis 3 sees the chapter once more while local recording is on.kiro_export.NOTICE_VERSIONstays 3.notice_showncheckskiro_telemetry_ack_version >= NOTICE_VERSION(or the terminal stamp, now written as 4), so a v3 acker keeps exporting after the bump.docs/system-specs/modules/metrics.mdnow says both versions are 4;docs/guides/telemetry-otlp-export.mdrows say "notice version 3 or later".Backwards compatibility
Compatible: a stored ack of 3 is still accepted by the Kiro telemetry export gate; the dashboard shows the chapter once more to v3 ackers whose recording is on, and an install off only by an old build's written
falsethat chose Keep off on v3 sees the off form once more, asoff_offer_duedocuments ("once the current notice is acknowledged").Removes nothing: the diff adds a string key, renders it and raises the notice version.
Tests
test/metrics/test_kiro_export_destination.py: newtest_a_v3_ack_still_exports_after_the_notice_moves_on(ack 3 with notice 4 still exports; fails ifNOTICE_VERSIONfollows the bump) andtest_an_ack_of_the_current_notice_exports(ack 4 exports). Tests that meant "the current notice was acknowledged" now writebeacon.PRIVACY_NOTICE_VERSION; the terminal stamp check matches whatnotice.pywrites.test/test_telemetry_consent_notice.py,website/src/test/App.test.tsx: version pins 3 -> 4.website/src/components/PrivacyChapter.test.tsx: asserts the new bullet renders.tsc -p tsconfig.app.jsonclean. black, flake8 and isort clean on the changed Python files.Proof: not provable -- other: a version bump plus copy; the new v3-export test was shown to fail by setting kiro_export.NOTICE_VERSION to 4 locally
Manual verification
Rendered the real SPA with fixture APIs (first-run Privacy chapter due) in Chromium at 1100x820 and confirmed the new bullet shows under "What is collected".
Screenshots / video
First-run Privacy chapter, light theme, with the new bullet third in "What is collected":
Related Issues
Refs #38145
Sibling sweep
Swept: git grep -nE "PRIVACY_NOTICE_VERSION =|NOTICE_VERSION =|collectedFeatures" -- src website/src
Sibling: src/kiro_crew/beacon.py -- changed here
Sibling: website/src/components/privacyNotice.ts -- changed here
Sibling: src/kiro_crew/metrics/kiro_export.py -- not affected -- its NOTICE_VERSION names the notice that first listed Kiro telemetry, which is still 3
Sibling: src/kiro_crew/metrics/notice.py -- not affected -- the terminal notice line lists no collected items, so it names no apps
Pattern harvest
Not generalizable: the notice copy and its version are product decisions; the bump rule is already written next to both version constants, and a reviewer lane enforced it here.
Checklist
Contribution License Agreement
Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.