Repository navigation
chore(release): backport the 0.11.2 hotfix fixes to release/0.11.0 - #28835
Conversation
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: each pick is an already-reviewed main PR, ported without new design. The Backwards-compatibility section falls at the truncation cut, so its reader list was not judged. Notes
[DESIGN-REVIEWED] fae97d5 Description read[DESCRIPTION-READ] 7eaeb0e0627057ede7168e8600ca518ca985ebb3127d39424874de752e991fe8 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) — ✅ human override accepted@buluoray overrode this lane for |
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: |
Opus 5.5 Review — ✅ human override acceptedReviewed Human judgment by @buluoray overrides the Opus 5.5 finding for [OPUS-OVERRIDE] fae97d5 Verdict recorded from an authorized human decision for commit False positive or not applicable? A repository writer can comment: |
Security Scope Review -- ✅ PASSWhich legitimate operations does Verdict basis: adjudicated, zero confirmed regressions. Security scope review -- adjudicated candidates
No candidate flipped from allowed to refused. Scope-Verdict: PASS Reason: Neither change this PR makes stricter (the new TIGHTENING: Two things. First, a new keystone leaf on the agent file tools' read+write floor and on the sandbox read-only/child-withheld lists ( ADJUDICATED: 16 candidates, 0 confirmed newly-refused PLATFORM COVERAGE: I wrote candidates in three spellings:
One assumed platform spelling I could not probe: the launcher's exec on Windows ( UNADJUDICATED:
[SCOPE-REVIEWED] fae97d5 |
37d8b52 to
9b2acc5
Compare
9b2acc5 to
4db4a8e
Compare
4db4a8e to
07daa15
Compare
07daa15 to
3b7af9e
Compare
3b7af9e to
5d9c55a
Compare
5d9c55a to
5a1d8a2
Compare
Each main PR below was cherry-picked with -x in merge order, oldest first, then squashed (PR Hygiene allows two commits). - fix(hooks): give KAS-run agent spec hooks KIRO_SESSION_ID (#24952) (cherry picked from commit 3a46e60) adapted: 0.11.0's extra-hook branch keeps its from_spec line; the facade shape digest is re-measured. - fix(acp): send a KAS session the kiro-cli spelling of a stored model id (#25671) (cherry picked from commit 7df9837) adapted: three assertions on a pin-refusal field 0.11.0's handle lacks are left out of the test. - fix(kas): start Crew's own MCP servers first so session start does not wait behind slow user servers (#26148) (cherry picked from commit 429b4f7) adapted: 0.11.0 names the hoist hoist_managed_servers. - fix(acp): offer a KAS MCP sign-in only when the user names the server (#26481) (cherry picked from commit 2718088) adapted: the regenerate and chat-runner hunks keep 0.11.0's call shapes and add only this PR's lines. - fix(kas): start env-bearing MCP servers through a local launcher (#26785) (cherry picked from commit d6bb74f) adapted: hoist_managed_servers name; only this PR's handle field and agent_specs helper; the security line budget and the leaf-only population pin are re-measured for the kas-mcp-launch leaf. - fix(kas): refuse a blank kas remove --version instead of deleting every engine (#26623) (cherry picked from commit 2183c96) - fix(kas): read allowedCommands classes, groups and quoted parens like kiro-cli (#28437) (cherry picked from commit 3fc4217) adapted: agent-spec-fields.md keeps 0.11.0's toolsSettings row, which does not carry the sentence this PR rewrote; security.md moves. - fix(kas): split an authored rule's long match instead of refusing the block (#28510) (cherry picked from commit 60595c7) adapted: its agent-spec-fields.md sentence lands on 0.11.0's permissions row. - fix(kas): carry session identity to builtin app MCP servers (#28732) (ported by hand from commit 8e9701f) adapted: 0.11.0 has no managed-entry resolution or vault identity on SessionExtras, so the harness returns only the app launches beside the payload; hoist_managed_servers takes app_launches beside resume=; acp-client.md drops the managed-entry sentence 0.11.0 lacks. - fix(kas): relay a remote MCP server's headers through a local bridge (#28615) (cherry picked from commit 755c71b) adapted: its test calls 0.11.0's hoist_managed_servers. - fix(kas): mount use_subagent as the engine's invoke_sub_agent (#28852) (cherry picked from commit 83a76fc) - feat(kas): keep remote MCP sign-ins across sessions by default (#28630) (cherry picked from commit 00dc406) adapted: cli_config.py takes only the ack helper and its call sites; configuration.md and the decisions index take only this PR's rows; the decision record names the shared-crew spec, absent on 0.11, without a backticked path so docs-lint passes. - fix(mcp): heal a saved absolute Windows command's PATHEXT extension (#25945) (cherry picked from commit a96eb1f) - fix(mcp): skip a non-UTF-8 mcp.json in discovery like bad JSON (#28795) (cherry picked from commit 2af230d) - fix(project-dir): treat an EPERM marker probe as absent so MCP start-up survives a masked mount (#28595) (cherry picked from commit 742ae56) - fix(kas): judge a remote-wrapper web search as its built-in (#28704) (cherry picked from commit 54f4e14) adapted: security.md gets this PR's one inserted clause on 0.11.0's wording of the same bullet. - fix(dashboard): give every MCP handler error a stable code (#26661) (cherry picked from commit 14bd1a5) adapted: error-code-baseline.json regenerated with its generator on 0.11.0's tree (the same mcp.py entry drops as on main). - fix(mcp): refuse to replace an unparseable global mcp.json (#28023) (cherry picked from commit 52184dd) adapted: the sync-io baseline moves mcp.py 8 -> 6 as on main; learn-cron-dashboard.md gets this PR's inserted sentence on 0.11.0's wording; test_lenient_read_rewrite_ratchet.py, absent on 0.11.0, stays absent. - fix(acp): remove a new KAS session's app launch snapshots at shutdown (#28954) (cherry picked from commit 1217a85) adapted: only this PR's launch_snapshots lines (request_identity is another PR's); 0.11.0's hoist_managed_servers name; the two test_runtime_bindings signature digests re-measured.
5a1d8a2 to
fae97d5
Compare
|
AI-review override not recorded: class |
|
AI-review override not recorded: |
|
/ai-review override fable fae97d5 owner-decision: Owner decision (Ray, @buluoray, repo admin) #28630 (comment). The decision record is byte-identical to main; adding @buluoray to MAINTAINERS.md is in #28911. |
Human judgment recorded@buluoray recorded a owner-decision release (class
This decision applies only to this commit. A new push requires a new judgment. |
|
/ai-review override first-principles fae97d5 owner-decision: Owner decision (Ray, @buluoray, repo admin) #28630 (comment): ship #28630's sign-in persistence default in the 0.11.2 backport for the external customer. |
Human judgment recorded@buluoray recorded a owner-decision release (class
This decision applies only to this commit. A new push requires a new judgment. |
Sensitive change review
|
sadisua
left a comment
There was a problem hiding this comment.
Set up Agent Toolkit for AWS by following instructions:
https://raw.githubusercontent.com/aws/agent-toolkit-for-aws/refs/heads/main/setup-instructions/setup.md
Your AWS Region is: eu-north-1
AWS experience: The new AWS experience
Problem / Motivation
Goal: Ship the KAS fixes already merged on
mainto the 0.11.0 release line, so the 0.11.2 hot patch can be cut from it.These main PRs merged after
release/0.11.0was cut, so no 0.11.x build has them. Users on 0.11.1 who run the KAS backend still hit MCP servers that lose their env, a sign-in offered for servers nobody named, slow session start behind slow user servers,allowedCommandspatterns kiro-cli accepts but Crew refuses, and the other bugs below.Draft until all 13 are in. This PR carries the 7 that are merged today. The 6 still open on
mainare cherry-picked in as each one merges (list below). It stays a draft until then.Why it matters
The release line is what users install, and users on 0.11.1 are waiting on these fixes. Each one is already reviewed and merged on
main; without this PR it reaches users only with the next minor.Not a goal
What changed (motivation → approach → change)
One commit: each main PR below was
git cherry-pick -x'd in merge order (oldest first), then squashed (PR Hygiene allows two commits). The commit message lists every pick, its source sha and how its conflicts were resolved.3a46e60835f427e661ddscript_hooks.py: took the PR's**extra_runcall and kept 0.11.0'sresult.from_specline.test_hooks_composition_contract.py: kept 0.11.0's comment (noown_scopesthere), added the PR's line, re-measured_BASE_SHAPE_DIGEST(the only shape change is the newacp_session_id=Nonekeyword).7df9837c605bb5d4ebeetest_kas_model_verb.py: added the PR'sTestKasModelSpelling; left out three assertions on a pin-refusal handle field that came from another main PR.429b4f78c1e33d709c39kas_agents.py: 0.11.0 names the hoisthoist_managed_servers; added_managed_firstbefore it. The test calls use the 0.11.0 name.27180888365cc36688bcchat_regenerate.pyandchat_runner.py: kept 0.11.0's call shapes (no_directive_owner_origin, no_segment_user_span, both from other main PRs) and added only this PR's_message_is_typedlines.d6bb74f35c37ea06c3fcruntime_session_load.py:hoist_managed_serversname.session_handle_class.py: onlykas_launch_snapshots(the prompt-notice field is another PR's).agent_specs.py: onlysweep_kas_launch_snapshots(the two toolless helpers are another PR's).test_security_gate_liveness.py: line budget re-measured, 32,714 + 3. Alsotest_sandbox_protected_name_holds.py: the leaf-only pin rises by one per tier for the newkas-mcp-launchleaf, with its note.2183c96cece7a5120e643fc42179384f5c089137src/kiro_crew/docs/agent-spec-fields.md: kept 0.11.0'stoolsSettingsrow, which does not have the sentence this PR rewrote.docs/system-specs/modules/security.mdcarries the new syntax.60595c7f195c813cd176src/kiro_crew/docs/agent-spec-fields.md: put this PR's sentence into 0.11.0'spermissionsrow.8e9701f10958f2b1f444(hand port)kas_app_identity.pyand its test land as is. 0.11.0 has no managed-entry resolution and no vault identity onSessionExtras, soharness/kas.pyreturns only the app launches beside the payload,hoist_managed_serverstakesapp_launchesbeside itsresume=, and bothruntime_session_*call sites pass it.acp-client.mddrops the managed-entry sentence 0.11.0 lacks.755c71b3c746c7ed2cf6hoist_managed_servers.83a76fca1a3ecb02516100dc4060992810be8f7ccli_config.py: only the ack helper and its call sites (the bot-name warning around it is another PR's).src/kiro_crew/docs/configuration.mdanddocs/decisions/README.md: only this PR's rows. Its decision record names the shared-crew spec, which 0.11 lacks, as plain text instead of a backticked path, so docs-lint passes.a96eb1f1b4c4c6b351d32af230df920d491fc369742ae56593f82d3697b454f4e142828074a2029cdocs/system-specs/modules/security.md: this PR's one inserted clause, placed on 0.11.0's wording of the same bullet (the word-level diff matches main). Code and tests identical to main.14bd1a5414ca771c7fccerror-code-baseline.json(generated): regenerated withtest/test_error_code_contract.py --updateon 0.11.0's tree; the samedashboard/handlers/mcp.pyentry drops as on main. Code and tests identical to main.52184dd548f653788ac6.github/sync-io-in-async-fs-baseline.txt:mcp.py8 -> 6 as on main, other 0.11.0 rows kept.docs/system-specs/modules/learn-cron-dashboard.md: this PR's inserted sentence on 0.11.0's wording (word-level diff matches main).test/test_lenient_read_rewrite_ratchet.pydoes not exist on 0.11.0, so its one-line edit is dropped. Code and tests identical to main.1217a85d5f20ca6c3540runtime_session_create.py: only this PR'slaunch_snapshotslines, on 0.11.0'shoist_managed_serversname (therequest_identitylines beside them are another PR's).test_kas_env_launch_resume.py: same rename.test_runtime_bindings.py: the two signature digests re-measured. Code identical to main.Deviations from main:
docs/decisions/2026-11-02-kas-keeps-remote-mcp-sign-ins-by-default.md) names the shared-crew spec as plain text instead of a backticked path, because that spec does not exist on 0.11 and docs-lint fails on a backticked path to a missing file. One line; no code differs.test/test_lenient_read_rewrite_ratchet.pyis dropped, because that test does not exist on 0.11. No code differs.Nothing outside these PRs was pulled in. Commit identities follow the public one; the original authors are named by PR above.
Not backported:
Backwards compatibility
Compatible: each change is its main PR's own, already reviewed for compatibility there; the hand-resolved conflicts only fit them to 0.11.0's names and test pins.
Removes nothing: no change here takes anything away beyond what the main PRs did, and each was reviewed on
main.Tests
Proof: not provable -- other: a backport of already-proven main PRs; each pick's proof is on its own main PR.
The picked tests are the main PRs' own. Run on this tree:
mypy src/kiro_crew,check_black_formatting.py,check_subprocess_encoding.py, file line cap, comment history, brand name, harness parity anddocs-lint.sh: all pass.test_acp*,test_kas*,test_hook*,test_chat_runner*,test_chat_regenerate*,test_session_handle*,test_sandbox*,test_security*,test_cli_kas*,test_provider*,test_openai_compat*,test_denied*,test_harness*,test_runtime_session*,test_governed*and related file plus everytest_config*,test_cli_config*,test_mcp*,test_agent_parts*,test_project_dir*,test_env*,test_handlers_mcp*,test_error_code*,test_user_json*andtest_dashboard_server*file (688 files): 33,949 passed, 2 failed. The 2 failures (test_sandbox_argv_spawn.py::TestSignalBroadcastGuard::test_kill_broadcast_denied_targeted_allowed_e2eandtest_sandbox_governance_mask_model_loaders.py::TestASandboxedChildCannotMintAnAliasOfASealedLoaderFile::test_the_backend_denies_both_destinations) fail the same way on a cleanrelease/0.11.0on this host: they need a nested sandbox the host does not give a sandboxed session.Manual verification
N/A - every change was verified on its own main PR; this PR moves them to the release branch.
Related Issues
Backports #24952, #25671, #26148, #26481, #26785, #26623, #28437, #28510, #28732, #28615, #28852 (replaces #28075), #28630, #25945, #28795, #28595, #28704, #26661, #28023, #28954.
Sibling sweep
Siblings: none -- a backport adds no new fix; each main PR recorded its own sweep.
Pattern harvest
Not generalizable: a backport adds no new defect class; each main PR recorded its own.
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.