Skip to content

chore(release): forward-port the 0.11.2 fixes to release/0.12.0 - #29148

Merged
buluoray merged 1 commit into
release/0.12.0from
crew/slack/forward-0.12.0-rc6
Nov 3, 2026
Merged

buluoray merged 1 commit into
release/0.12.0from
crew/slack/forward-0.12.0-rc6

Conversation

@buluoray

@buluoray buluoray commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Bring every fix the 0.11.2 hotfix carries, and that release/0.12.0 still lacks, onto release/0.12.0 so v0.12.0-rc.6 ships them and no 0.11.2 fix regresses when an insider moves to 0.12.

release/0.12.0 was cut before these main PRs merged. The 0.11.2 backport (#28835) put them on release/0.11.0, so without this PR an insider on 0.12.0-rc.6 would lack fixes a 0.11.2 user has. #28925 merged on main after #28835 and is included so the two lines stay level.

Why it matters

The KAS and MCP failures these PRs fix (spec hooks that silently do nothing, an unparseable mcp.json replaced, remote MCP sign-ins asked for again in every session, MCP start-up dying on a masked mount, app MCP servers without session identity) would hit every insider on the 0.12 line. A user who upgrades from 0.11.2 to 0.12.0-rc.6 would see them come back.

Not a goal

What changed (motivation → approach → change)

Each main PR was cherry-picked with -x onto origin/release/0.12.0 @ 44ce6619c8 (v0.12.0-rc.5), in main merge order, oldest first, then squashed into one commit (PR Hygiene allows two). The commit message lists every source commit.

# PR main commit Pick
1 #28023 refuse to replace an unparseable global mcp.json 52184dd548 conflict in surrounding lines only
2 #28437 read allowedCommands classes, groups and quoted parens like kiro-cli 3fc4217938 conflict in surrounding lines only
3 #26681 answer /context, /usage, /help and /agent locally on KAS 84ea89db7c clean
4 #28795 skip a non-UTF-8 mcp.json in discovery like bad JSON 2af230df92 clean
5 #28510 split an authored rule's long match instead of refusing the block 60595c7f19 clean
6 #28732 carry session identity to builtin app MCP servers 8e9701f109 clean
7 #28615 relay a remote MCP server's headers through a local bridge 755c71b3c7 clean
8 #28852 mount use_subagent as the engine's invoke_sub_agent 83a76fca1a clean
9 #28630 keep remote MCP sign-ins across sessions by default 00dc406099 conflict in surrounding lines only
10 #28595 treat an EPERM marker probe as absent so MCP start-up survives a masked mount 742ae56593 clean
11 #28704 judge a remote-wrapper web search as its built-in 54f4e14282 clean
12 #28954 remove a new KAS session's app launch snapshots at shutdown 1217a85d5f clean
13 #28925 run spec hook commands longer than 4096 characters on kas 0fe72d6165 adapted (doc only)

For every pick except #28925, the added and removed lines match its main commit line for line. The conflicts were all in neighbouring lines that release/0.12.0 has a different version of:

Backwards compatibility

Compatible: each change is the one already merged on main and on release/0.11.0 (except #28925, which is on main only). Each PR's own compatibility note applies unchanged.

Removes nothing: no pick hides, deletes or renames anything 0.12.0 accepts. #28630 changes the agent.kas_persist_mcp_sign_in default, recorded in docs/decisions/2026-11-02-kas-keeps-remote-mcp-sign-ins-by-default.md (shipped by #28630).

Tests

Proof: not provable -- other: a forward-port of main commits whose proofs ran on their own PRs

  • The 26 test files this diff adds or changes: 2070 passed.
  • Each pick's own tests were run after it applied, before the squash.
  • test/test_acp_copilot_backend.py fails 3 tests on this host, and the same 3 fail on clean origin/release/0.12.0, so they come from the host and not from this diff.
  • Gate on the changed files: flake8, isort, black (py310), mypy src/kiro_crew, check_file_line_cap, check_harness_parity, check_comment_history, check_brand_name, check_subprocess_encoding, check_sync_io_in_async and docs-lint all pass. check_black_formatting.py could not run here because the local Black is 26.3.1 and the script needs 26.10.1. CI runs it.

Manual verification

N/A - every pick was verified on its own main PR. The rc.6 insider build is the end-to-end check.

Related Issues

Forward-port of the 0.11.2 hot patch (#28835) to the 0.12 line, ahead of v0.12.0-rc.6.

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

Contribution License Agreement

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

Every main PR the 0.11.2 backport carried that release/0.12.0 still
lacks, plus #28925, cherry-picked with -x in main merge order, oldest
first, then squashed (PR Hygiene allows two commits). Each pick's
added and removed lines match its main commit except where noted.

- fix(mcp): refuse to replace an unparseable global mcp.json (#28023)
  (cherry picked from commit 52184dd)
  conflict in surrounding lines only; the changed lines are main's.
- fix(kas): read allowedCommands classes, groups and quoted parens like
  kiro-cli (#28437)
  (cherry picked from commit 3fc4217)
  conflict in surrounding lines only; the changed lines are main's.
- feat(dashboard): answer /context, /usage, /help and /agent locally on
  KAS (#26681)
  (cherry picked from commit 84ea89d)
- fix(mcp): skip a non-UTF-8 mcp.json in discovery like bad JSON
  (#28795)
  (cherry picked from commit 2af230d)
- fix(kas): split an authored rule's long match instead of refusing the
  block (#28510)
  (cherry picked from commit 60595c7)
- fix(kas): carry session identity to builtin app MCP servers (#28732)
  (cherry picked from commit 8e9701f)
- fix(kas): relay a remote MCP server's headers through a local bridge
  (#28615)
  (cherry picked from commit 755c71b)
- 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)
  conflict in surrounding lines only: docs/decisions/README.md and
  configuration.md lack neighbouring rows 0.12.0 does not carry.
- 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)
- fix(acp): remove a new KAS session's app launch snapshots at shutdown
  (#28954)
  (cherry picked from commit 1217a85)
- fix: run spec hook commands longer than 4096 characters on kas
  (#28925)
  (cherry picked from commit 0fe72d6)
  adapted: agent-host-contract.md gets main's inserted sentence in
  0.12.0's paragraph, which lacks main's later user-hook-files text.
@buluoray
buluoray requested review from a team and KiroCrewCaptain as code owners November 3, 2026 12:06
@buluoray
buluoray requested review from chenmingwei23 and removed request for a team November 3, 2026 12:06
@buluoray
buluoray merged commit 5ca4baf into release/0.12.0 Nov 3, 2026
18 of 20 checks passed
@buluoray
buluoray deleted the crew/slack/forward-0.12.0-rc6 branch November 3, 2026 12:06
@github-actions

github-actions Bot commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

Sensitive Change Review

Sensitive files changed:

  • src/kiro_crew/hook_runtime/gate_tiers.py

  • No reasoning template for c33250d yet.

How to clear this check:

  1. The PR author posts the template below as a PR comment, with
    every field filled (N/A needs a reason). Keep the
    Head commit line: it must name the current head SHA.
  2. The check turns green within a minute, with no push.
  3. A new push resets this check: post it again for the new head.

Only the author's comment counts this way. A reviewer can instead
paste the filled template into an "Approve" review on this head.
Merging still needs a reviewer's approval.

## Sensitive change review
- Head commit: c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa
- What rule changed:
- Before -> after (what was allowed/blocked, now):
- Worst case if wrong (breaks commands / leaks secret / locks user out):
- Evidence checked (which sandbox result proves it):
- No regression:
  - Tools still work (which tools/commands were run, result):
  - Backward compatible (old config / old data / old callers still work, how checked):
  - Existing tests (which suites passed on this head):
- How to undo:

@github-actions

github-actions Bot commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ⏭️ skipped

Revision c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

@github-actions

github-actions Bot commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: each pick forward-ports a fix already merged on main. The one default flip (kas_persist_mcp_sign_in) and the "off by default" pin it removes are backed by the recorded owner decision that #28630 ships.

Notes

[DESIGN-REVIEWED] c33250d

Description read

[DESCRIPTION-READ] 3119b7556a040cbed0c09427422383efa99d5e3286815c162f9259cc3f593f3a

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 Nov 3, 2026

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — 🔴 BLOCK

Premise-level review of c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa — 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 does not.

First-Principles-Verdict: BLOCK

Item 9 changes a default that nearly everyone gets. agent.kas_persist_mcp_sign_in goes from false to true, and a stored false is rewritten to true once. The base RFC list has no non-draft RFC that covers this.

What this change ships

Inventory (10 items) — 9 justified

Intent: give 0.12 insiders every fix that 0.11.2 already shipped, so nothing regresses when they upgrade. This is a FIX (a forward-port).

  1. Dashboard refuses to overwrite an unparseable global mcp.json — justified
  2. allowedCommands classes, groups and quoted parens read the way kiro-cli reads them — justified
  3. /context, /usage, /help and /agent are answered locally on KAS — justified
  4. A non-UTF-8 mcp.json is skipped in discovery, the same as bad JSON — justified
  5. A long match in an authored rule is split instead of the block being refused — justified
  6. Builtin app MCP servers receive session identity — justified
  7. A remote MCP server's headers are relayed through a local bridge — justified
  8. use_subagent is mounted as the engine's invoke_sub_agent — justified
  9. Remote MCP sign-ins persist across sessions by default, and a stored false is migrated to true once — rides along: a default flip with no RFC on the base
  10. EPERM marker probe, web-search classification, snapshot cleanup, and spec hooks over 4096 characters — justified

Blockers

product-shape change without accepted RFC: KAS keeps remote MCP sign-ins by default — rides along
Trigger: config-baseline.json changes "kas_persist_mcp_sign_in": false to true, and superseded_defaults adds | agent.kas_persist_mcp_sign_in | false -> true, #28630 | adopts |. The only record of the decision is docs/decisions/2026-11-02-kas-keeps-remote-mcp-sign-ins-by-default.md, a file this diff adds, and it is not an RFC. The base RFC status list has no non-draft RFC for KAS sign-in persistence; the MCP RFCs it lists are draft, except rfc-owner-written-mcp-auto-approve (implemented), which is about auto-approve, not sign-in persistence.
Outcome: every 0.12 insider gets a gateway-wide credential vault turned on, and a false they stored gets overwritten. No decision this lane accepts is on record.
Clears when: an RFC covering KAS's default persistence of remote MCP sign-ins is merged non-draft on the base, followed by a push or rebase; or a maintainer posts /ai-review override first-principles c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa product-shape: <decision>.

[FIRST-PRINCIPLES-REVIEWED] c33250d

Description read

[DESCRIPTION-READ] c16d2e16840cb84e1f208f6e15b9bf1d8ad51d87fd8893b49bdc62d26c99ecf2

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.

buluoray added a commit that referenced this pull request Nov 3, 2026
Declare 0.12.0-rc.6 in all three version files so the next insider
RC ships the 0.11.2 fixes forward-ported by #29148.
@github-actions

github-actions Bot commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] c33250d

False positive or not applicable? A repository writer can comment:
/ai-review override gpt c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] c33250d

Verdict parsed from the review's SHA-scoped output markers for commit c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa.

False positive or not applicable? A repository writer can comment:
/ai-review override fable c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

Security Scope Review -- ✅ PASS

Which legitimate operations does c33250d4d3fbbf1d09d7cbc12d0a5b5ff0a754aa newly refuse? Updated in place on each push. A script-confirmed regression blocks; the model proposes candidates and never decides a refusal.

Verdict basis: adjudicated, zero confirmed regressions.

Security scope review -- adjudicated candidates

  • macos-latest (platform posix): 10 classified, 0 newly refused, 10 still allowed.
  • ubuntu-latest (platform posix): 10 classified, 0 newly refused, 10 still allowed.
  • windows-latest (platform windows): 6 classified, 0 newly refused, 6 still allowed.

No candidate flipped from allowed to refused.

Scope-Verdict: PASS

Reason: Nothing in this diff makes the shell deny composite (sensitive-bash, exfil, deny-rules) stricter. The only shell-facing change is in platform/allowed_commands.py, and it widens the spec allowedCommands grant tier: entries may now use classes, groups and alternation, and commands with parentheses inside quotes become grantable. The narrowings that come with it (ASCII and no backslash when quoted grouping is present, the \</\> escape refusal, a shared step budget) only touch entries or commands that were refused or never matched before. Two guards outside the classifier have no verdict and are named below.

TIGHTENING: allowed_commands.parse/plain_command (spec allowedCommands grant reader) -- the one threat it refuses is grouping characters or regex syntax that could widen a grant past what kiro-cli grants: + if not command.isascii() or "\\" in command: / + return False in _grouping_only_quoted, together with + if not nxt.isascii() or not nxt.isprintable() or nxt.isalnum() or nxt in " <>": / + raise _Refused in _Parser._escape. Before this diff every such command or entry was already refused: UNGRANTABLE_COMMAND_CHARS held (){}, and any metacharacter other than .* returned None. The literal-plus-.* reader (_literal_segments) is unchanged, and it uses none of the shared budget.

ADJUDICATED: 16 candidates, 0 confirmed newly-refused

PLATFORM COVERAGE: posix and windows. The diff assumes one thing about platforms: _grouping_only_quoted says single- and double-quoted ( ) { } read as plain text in POSIX shells, zsh and PowerShell. Every quoted-grouping candidate is written as both a posix row and a windows row (git commit -m "fix(kas): ...", git commit -m 'feat(dashboard): ...'). Other candidates cover the .exe/.cmd executable names (npm.cmd, .venv\Scripts\python.exe), \ separators, and a quoted PowerShell Get-ChildItem -Filter "*.py". No row is any. Not probed: typographic quotes on PowerShell. The diff refuses non-ASCII commands that contain grouping characters, and I found no legitimate operation that needs both.

UNADJUDICATED:

  • KAS permission-block splitting (acp/kas_permissions.py, parse_user_permissions / _parse_rule / merge_user_permissions). Call path: agent spec permissions → merge_user_permissions → parse_user_permissions → _parse_rule. A new refusal fires when the rule count after splitting exceeds _MAX_RULES. Any rule that would split was refused before (a match over 64 patterns refused the whole block), so no block that used to travel should now be refused. Inputs I would have submitted:

    • a block of 200 rules where one has a 65-pattern match;
    • a 64-pattern allow with a 65-pattern exclude beside a deny.

    These are not shell command lines, so there is no verdict.

  • KAS identity-disagreement exemption (permission_requests._is_kas_remote_builtin, in build_permission_event). This only relaxes a guard: a serverName: "remote" frame with toolId remote_<tool> that is a mapped built-in is no longer marked mcp_identity_unreadable. Nothing becomes newly refused, but the guard is not shell-classifiable. Inputs I would have submitted: a KAS web_search permission frame with {serverName:"remote", toolName:"web_search"} and toolId remote_web_search, and the same frame after a tool_call that named a real server.

  • KAS slash-command routing (kas_local_commands.answers_locally). On KAS, /context <arg> and /agent <word> (other than list) now fall through to the "doesn't work with KAS" notice. This is dashboard routing, not a security guard, and it is not classifiable. Inputs: /agent list, /agent reviewer, /context, /context add docs/.

[SCOPE-REVIEWED] c33250d

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.

1 participant