Repository navigation
fix: backport eight MCP, Bot and KAS fixes to 0.12.1 (batch 27) - #32155
Conversation
(cherry picked from commit 6f69f2a)
The mate editor lost its Session color field when the New editor replaced the classic dialog, so a mate's color could only be picked at create time. useCrewEditor still loads, dirty-tracks and saves session_color, so render the shared SessionColorField on General under Display name. Tests cover changing and clearing the stored color. Closes #29823 (cherry picked from commit ab6df81)
…31573) When the credential gate refuses a minted approval URL, the mint ends with reason mint_url_rejected and, when it can, a copy-ready rejected_endpoint. The Connections card already turns that into the endpoint to list in oauth_endpoints.json plus a link to the guide; the MCP Servers row ignored both and always said "Could not start the sign-in. Try again.", a retry that can never clear the refusal. The row now shows the same copy, remedy line and guide link as the card. Its error renders through ErrorNotice, whose agent hand-off is withheld while the MCP table holds unapplied edits, as MCP Management already does. The guide URL moves to utils/docsLink so both surfaces share it, and the guide and security spec name the row. Closes #27033 (cherry picked from commit f52b36b)
…1476) When the process Kiro Crew spawned for a kiro-cli session is not the one that reads its stdin (a wrapper or launcher sits in front of kiro-cli) and it dies before its tree does, for example killed during the initialize handshake, the descendants keep the write ends of stdout and stderr. asyncio's subprocess transport then never sees every pipe close, never finishes, and never closes stdin, so kiro-cli reads no EOF. On macOS no member of the leaderless group can vouch for it and no descendant was recorded yet, so nothing signals the tree either: kiro-cli, its engine and any MCP server it had started keep running after the session is closed, for the life of the gateway. AcpRuntime's POSIX kill now aborts its own write end of the root's stdin before it drops the handle. It signals nothing: only holders of the pipe's other end, which are this spawn's own tree, see the EOF. It uses the stdin pipe's abort rather than close, which waits for buffered bytes to drain, and never the subprocess transport's close, which kills a root that is still running. Part of #19141 (cherry picked from commit 15ec103)
…31797) A patrol loop's max_cycles counts its whole life and the server caps it at 1000, so at the 90 s interval one loop lasts about 25 hours, renewals included, and the skill gave no way past it. A wake cannot re-arm its own loop: while it runs monitor_start is refused, and once the wake stops it the server refuses the replacement, which leaves no patrol at all. The arm step and the standing instruction now have the conductor ask in its reply on the "10% or less left" budget line while work is live, and re-arm (autonudge_stop, then monitor_start with the same interval, bounds and instruction) in the turn the person's own message starts. The spec states the limit as a failure mode and why the work-ledger watch's bound extension is not used, and the contract test pins the clause and the server's budget-line wording. (cherry picked from commit 492f614)
(cherry picked from commit 1c333b9)
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 screenshots. This backport changes three dashboard surfaces, the PR attaches no image, and the blind read was not performed. Evidence gaps
Notes
[UX-REVIEWED] 1e0b7ba Description read[DESCRIPTION-READ] f123198a112076a8f62f6358856e99e3a7a42f7427b35e8d60cfdc48f02240ff 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 |
Design Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)Design-level review of Design-Verdict: PASS Cannot evaluate: missing a screenshot of the new mate-editor Session color field and the reshaped MCP-row sign-in error. Windows verification is also missing for the two Windows-only picks, and no CI run on this branch covers them. BlockersMissing evidence for this revision (UI and Windows) [DESIGN-REVIEWED] 1e0b7ba Description read[DESCRIPTION-READ] 11896bb931fb93ef20034e5ab9ac34bd52bea9daf1db3e8cfd65f25461b601c7 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 Each item is an unchanged cherry-pick of a fix already merged on Notes
What this change shipsInventory (8 items) — 7 justifiedIntent: ship eight fixes already merged on main to 0.12.x users in the 0.12.1 patch. This is a FIX (a backport).
[FIRST-PRINCIPLES-REVIEWED] 1e0b7ba Description read[DESCRIPTION-READ] 1833f487dad1033077c19a238e75e2a6ccc579ed248dc37ff57e315018e2359f 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: |
Problem / Motivation
Goal: Bring eight small MCP, Bot and KAS fixes that merged on
mainafterv0.12.0ontorelease/0.12.0, one commit per fix, so the 0.12.1 patch ships them and each one can be reverted alone.Why it matters
PATHEXTcasing repair, so its server command can fail to resolve.work_reportafter it is bound is refused withcrew_log_unit_unknown.pyright-langserverpile up because a dead runtime's kiro-cli tree does not exit when a wrapper sits in front of it.Not a goal
main; nothing is changed only here.CHANGELOG.mdchange. Cuttingv0.12.1-rc.Nand the bare tag follow docs/build/release.md and stay with the maintainers.kas_permissions.py, next to the permission gate), fix(mcp-gateway): do not wrap a server a config scope mutes #31513 (+354 across 8 files), fix(mcp-shared): end the tool-policy failure window when the gateway answers again #30907 (+330, shared MCP dispatch), fix(mcp): name an mcp env placeholder that will not resolve #31562 (+372, touchesacp/launch.py), fix: keep a gated managed server out of app agent specs #31337 (12 files across the app bridges), perf(crew-log): commit a session's queued appends with one write and one fsync #31568 (+828, crew-log write path).What changed (motivation → approach → change)
Each main squash commit was cherry-picked with
-xontorelease/0.12.0@4d9f6b123e, in merge order, one commit each, keeping its original author.6f69f2a5fb7aaf083bc2c50773c8b671b5f2eda7ab6df81cec280459d168f52b36b86ea019d27b96a13dc00715f0efd13b4215ec10364713990b51ae492f6149d21beb1acf5d1c333b964b1e0b7bae98No hunk was resolved by hand. Every pick's added and removed lines are identical to its main commit, and every pick carries a test. #31797 changes only the
pipeline-conductorskill text plus its test. #31820 is a small edit inagent_parts/spec_build.py; #31476 adds one step toacp/runtime_tree_kill.py.builder-security checked: no pick changes a bind, a listener, a route, an auth check, an ingress rule or a storage grant.
Backwards compatibility
Compatible: each change is the one already merged on
main, and each PR's own compatibility note applies unchanged.Removes nothing: no pick hides, deletes or renames anything 0.12.0 accepts.
Tests
Proof: not provable -- other: a backport of main commits whose proofs ran on their own PRs
release/*, so the related set was run here: every touched test file plus every test file that imports a touched module, 73 files, one pytest process per batch of 20,-n 0. 5171 passed, none failed.Manual verification
N/A - every pick was verified on its own main PR. The 0.12.1 RC soak is the end-to-end check.
Screenshots / video
Why no screenshot: each frontend change here is applied from
mainunchanged; its look is reviewed on its own PR, not here.Related Issues
Backport to the 0.12 line for 0.12.1 of #31820, #31846, #31999, #31573, #31976, #31476, #31797 and #32111. Source issues: #25717, #30311, #29823, #27033, #30352, #19141, #20870, #27799.
Sibling sweep
Siblings: none -- each source PR carried its own sibling sweep; this PR only carries those fixes.
Pattern harvest
Not generalizable: this PR only carries fixes already on
main. Each source PR has its own harvest.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.