Repository navigation
feat(apps): pin issue-radar and auto-improvement for adoption - #36446
Conversation
|
Catalog rows: kirodotdev/KiroCrewApps#77 switches the |
7283a32 to
28398aa
Compare
|
Rebased: main moved in docs/system-specs/modules/history.md since this head went green (6149be8 -> d4b588b); CI re-running. The previous head 7283a32 had its review lanes, Dependency License Gate and readiness red on the repo-wide installation-token rate limit ( |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: this adds two pin rows through the existing adoption machinery, and the wheel copies remain as the fallback. Notes
[DESIGN-REVIEWED] 8e5f2a9 Description read[DESCRIPTION-READ] dbc2f71b22a7863c7ef2e7ca7fd475cd24a50fdccb45fe9bb2a0e3507f030512 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 |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks. Both rows reuse the existing adoption-pin mechanism to carry out the AGENTS.md "No built-in apps" move-out under #26967. What this change shipsInventory (5 items) — 5 justifiedIntent: let hosts running Issue Radar or Auto-Improvement as built-ins move onto each app's own repository at the next restart, keeping their data. This is an ADDITION.
[FIRST-PRINCIPLES-REVIEWED] 8e5f2a9 Description read[DESCRIPTION-READ] c939e6e9bac117ddacb0f267daa69de46ddb544b3d2190bcc4ca67b78ac8db08 That sha256 names the PR description this verdict read. 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: |
|
Backend Lint & Type Check (3.12) (and the Coverage Gate that waits on it) is red on main, not on this diff: |
Code change approver recommendation: ApproveWhy: the decider estimates a 10% chance that this change introduces a defect, within the 12% limit for approving on its own, and every check passes. This PR only changes data. It adds two rows to the adoption pin table, issue-radar 0.1.1 and auto-improvement 1.0.2, in the same shape as the five apps already pinned. It moves each wheel app.json version up to match its pin, which the version-equality rule (D12) requires and test_every_shipped_pin_matches_its_wheel_built_in_version enforces. It also updates the shipped-table test (_SHIPPED_PINS) and the two module specs. The table now has 7 rows, well under the 64-row cap (MAX_PINS). Core code has no imports of either app's built-in module and no app-name string literals outside the wheel copies, so adoption can't leave core seams calling unloaded modules. That problem is why aws-control was held back. On a host, every adoption check runs again at restart and fails closed when it doesn't hold: the tree hash, identity and version, the MCP server shape, permission parity and the SDK level. If a check refuses, the host keeps the wheel copy. That limits the risk to mistakes in the pin values themselves, which can't be checked offline. The PR reports a network Several CI checks hadn't finished at deci 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 |
|
Holding the rebase push: the auto-improvement pin is missing a security fix that landed on main today. Rebasing onto The pinned tree Proposed fix, which keeps the frozen goal:
The other way out is to hold auto-improvement back from this PR. That changes the frozen Goal, so only a maintainer can choose it. The current head |
28398aa to
c993367
Compare
|
Hold resolved: the auto-improvement pin now carries #30602.
Auto-merge stays armed. The KiroCrewApps catalog row moves to the same commit in kirodotdev/KiroCrewApps#77. |
Flips both first-party apps from a builtin source to a git source in KiroCrewCaptain, each pinned at the commit the Kiro Crew adoption pin table names for it (kirodotdev/KiroCrew#36446). The Kiro Crew org row (issue-radar) gains the KiroCrewCaptain author URL; auto-improvement keeps its person attribution.
c993367 to
f57cab5
Compare
|
Rebased: main moved in test/test_builtin_adoption_pins.py (#36344, #36356) and the adoption checks (#36325) since this head went green (6d61070 -> 02bad16); CI re-running. The old head c993367 had no review finding; its only reds (Analyze (actions), Dependency License Gate, GPT 6.1 Review, Publish readiness signal) were the installation-token and Bedrock rate limits. On the rebased tree |
f57cab5 to
a3e63c6
Compare
|
Rebased onto
Re-verified on the new base:
|
|
Head a3e63c6 has one remaining red, and it comes from
#37005 added |
|
Hold: do not rebase-and-push this PR until the issue-radar pin is recomputed. The shard 2 red on
If this lands as it is, every host adopts issue-radar at its next restart and runs the What has to happen first (outside this PR's branch):
After that, the next round rebases onto main, re-runs |
Adds both former built-ins to the wheel's adoption pin table, each at the main head of its own KiroCrewCaptain repository (CI green there), with the row scripts/adoption_pin.py compute printed and verify re-checked. A host whose record is builtin-owned swaps it in place onto that tree at the next restart, keeping data, secret, grants and enabled. Each wheel built-in app.json version moves to the pinned version, since a first-adoption pin must equal it (D12). Both pins need App Kit level 2: issue-radar for ctx.auto_approve and agentTools, auto-improvement for ctx.lifecycle. Part of #26967
a3e63c6 to
8e5f2a9
Compare
|
Hold resolved. The issue-radar pin is recomputed, and auto-improvement was re-checked the same way.
Both repositories' PR and The branch is rebased onto |
The Kiro Crew pin PR (kirodotdev/KiroCrew#36446) now pins issue-radar 0.1.2 at KiroCrewCaptain/kirocrew-app-issue-radar@8e57db85e15e and auto-improvement 1.0.3 at KiroCrewCaptain/kirocrew-app-auto-improvement@0e1a22d2e10b, each carrying the wheel-copy fixes that landed after the earlier commits. Each source ref moves to the commit its pin row names.
|
Green-age check on head 8e5f2a9 (no push):
The green on this head still describes today's base for this PR's closure, so it is left un-rebased (re-sync budget kept at 2/3). |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Approved. On this head the review reproduced the archiveSha256 and treeSha256 of both pinned trees (issue-radar at 8e57db85, auto-improvement at 0e1a22d2). It confirmed that each pin version equals the pinned manifest and the bumped wheel app.json, which test_every_shipped_pin_matches_its_wheel_built_in_version pins. It also checked that permissions, agentTools, agents, skills and the MCP autoApprove list match the wheel copies, and that the runner gate, may_arm_auto_merge and the owner gates carry the wheel-copy safety fixes behind host shims that fail closed.
Done well: you held the push twice when a pinned tree was missing a wheel-copy safety fix (#30602, then #31713 and #32503), ported the fixes, and recomputed both pins.
Optional:
- Before merge, run the drift check again (git diff b81cf5a origin/main over both wheel-copy trees), because open PRs such as #30251 still edit these wheel copies. A short PR comment could also tell those authors that an
App repo:line will be required once this lands. - The "Held back" list in the body is stale. #36321 and #36322 are already in this PR's base, and #37411 now pins auto-research and project-scaffolder. A body edit is enough.
- docs/system-specs/modules/issue-radar.md:103 says the rail keeps the "Radar" glyph, but "Radar" is not a registered nav glyph, so the rail shows the app's own icon. Users see no change. Only the wording is off.
…es (#77) * feat: list issue-radar and auto-improvement from their own repositories Flips both first-party apps from a builtin source to a git source in KiroCrewCaptain, each pinned at the commit the Kiro Crew adoption pin table names for it (kirodotdev/KiroCrew#36446). The Kiro Crew org row (issue-radar) gains the KiroCrewCaptain author URL; auto-improvement keeps its person attribution. * chore: move issue-radar and auto-improvement to their new pinned commits The Kiro Crew pin PR (kirodotdev/KiroCrew#36446) now pins issue-radar 0.1.2 at KiroCrewCaptain/kirocrew-app-issue-radar@8e57db85e15e and auto-improvement 1.0.3 at KiroCrewCaptain/kirocrew-app-auto-improvement@0e1a22d2e10b, each carrying the wheel-copy fixes that landed after the earlier commits. Each source ref moves to the commit its pin row names.
Problem / Motivation
Goal: A host that has Issue Radar or Auto-Improvement as a built-in moves onto the app's own repository at its next restart, keeping everything it had.
Both apps now live in their own repositories (KiroCrewCaptain/kirocrew-app-issue-radar and KiroCrewCaptain/kirocrew-app-auto-improvement), ported up to the current wheel copy, with green CI on ubuntu, macOS and Windows. The wheel's adoption pin table does not name them yet, so no host adopts them and every fix keeps having to land twice.
Why it matters
Until a pin row exists, these two apps cannot leave core: every host keeps running the wheel copy, and the removal of the built-ins in #26967 cannot start for them.
Not a goal
src/kiro_crew/apps/builtins/<app>/,website/src/apps/<app>/); only theirapp.jsonversion moves.What changed (motivation → approach → change)
Each row is the output of
scripts/adoption_pin.py compute, copied intoadoption-pins.jsonand the_SHIPPED_PINStable:minAppSdkLevel8e57db85e15ectx.auto_approve,agentTools,workerAutoApprove)0e1a22d2e10bctx.lifecycle)Each wheel built-in
app.jsonversion moves to the pinned version: a first-adoption pin must equal it (D12), andcomputerefuses the row withversion_mismatchuntil it does. Each repository'sapp.jsondeclares the samepermissionsas the wheel copy (the same threeagentToolsfor issue-radar; the same agents, skills and MCP server name andautoApprovelist for auto-improvement), so E9 holds nothing back. The two module specs record the adoption.Each pinned tree carries every runtime change the wheel copy has on
mainb81cf5a3acc0:may_arm_auto_merge, never arming auto-merge on GitLab, with the editor's locked toggle and its strings), fix(dashboard): give every bad-body 400 a code #32503 (a code on every bad-body 400), fix(i18n): write the generic noun crew in each script in bn, ja, ko, zh-CN, ru #37301 (the pipeline strings' native crew noun) and the test-only ErrorNotice impact levels: build the accepted RFC (component, then call sites by area) #37262, test: check real outcomes in nine new weak tests #36562, test: use synthetic credentials in redaction test fixtures #31863 and fix(crew): saving a Bot resets the model and effort its chat picked #36062.~/.kiro. test: use synthetic credentials in redaction test fixtures #31863 and test: drop the stale location claim from the runner coverage docstring #35994 change only test files and are not ported.Held back from the batch that reported ready:
campaigns.dbstill besideapp.jsonon a host that skipped fix(research-lab): keep the campaigns database in the app data dir #32650, andtest_auto_research_is_not_pinned_while_the_swap_can_retire_its_old_databaserefuses the row on main.App repo: KiroCrewCaptain/kirocrew-app-issue-radar@8e57db8
App repo: KiroCrewCaptain/kirocrew-app-auto-improvement@0e1a22d
Backwards compatibility
Compatible: two rows are added and two versions move up (issue-radar 0.1.0 to 0.1.2, auto-improvement 1.0.0 to 1.0.3). A host whose record is builtin-owned adopts at the next restart and keeps
data/,.app_secret, the grants andenabled; a host that cannot adopt keeps the wheel copy unchanged. No other row changes.Removes nothing: the wheel copies, their routes and every existing pin row stay.
Tests
test/test_builtin_adoption_pins.py::test_the_shipped_pin_file_validates_and_pins_exactly_the_shipped_tabletakes the two new rows;test_every_shipped_pin_matches_its_wheel_built_in_versionholds the version bumps.Proof: PROVEN test/test_builtin_adoption_pins.py::test_the_shipped_pin_file_validates_and_pins_exactly_the_shipped_table (head 8e5f2a9)
Determinism: no test logic changes, only the pinned data table. Every
test/test_builtin_adoption*.pyandtest/test_core_auto_improvement_decoupling.pyonorigin/mainb81cf5a plus this commit: 361 passed, 1 skipped. Black (at the pinned 26.10.1), flake8, isort, the comment-history, brand-name and line-cap gates pass; the diff has no frontend change.Manual verification
scripts/adoption_pin.py verifyover the network onorigin/mainb81cf5a plus this commit, with a throwawayHOMEandKIROCREW_HOME: all seven rowsokunder main's current tree and seam-level checks. Each pinned commit is the head of its repository'smain, and its pull request andmaincheck runs are all green (Tests on ubuntu, macOS and Windows, the reproducible UI bundle, README, internal-content-scan with credentials present and every added line checked). Each repository's port was a three-way merge of the wheel copy's changes since its last sync, with every catalog value checked against the base wording first.Related Issues
Part of #26967
Catalog rows: the kirodotdev/KiroCrewApps PR linked in a comment below switches both
sourcerows to the same commits. It is not auto-merged: the driver merges it right after this PR lands.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)