Skip to content

chore(release): cut 0.11.1-rc.2 with four more hotfix fixes - #26730

Merged
bolichen97 merged 1 commit into
release/0.11.0from
backport/0.11.1-rc2
Oct 31, 2026
Merged

bolichen97 merged 1 commit into
release/0.11.0from
backport/0.11.1-rc2

Conversation

@bolichen97

@bolichen97 bolichen97 commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Cut 0.11.1-rc.2 on release/0.11.0 with four more fixes from main, so the next hotfix RC carries them.

The 0.11.1 line (rc.1, #26648) is missing four fixes that are merged on main: the CLI installers do not download Crew's built-in KAS engine (it stays opt-in through agent.kas_bundle), quitting the desktop app during start can leave a gateway running, some background tasks can be dropped before they finish, and one security fix: fix(kas): refuse a file write a fallbackAction deny spec does not cover.

Why it matters

These fixes reach users only through the release line. A one-line install on 0.11.1-rc.1 has no built-in engine until the user runs kirocrew kas install by hand.

Not a goal

What changed (motivation → approach → change)

One commit; its message lists every pick and its source sha.

Product shape: #26384 makes the CLI installers download the pinned KAS engine by default (--no-kas opts out). That is an owner decision recorded on main: Zezhen, 2026-10-30, #26384 (comment), carried on main by a maintainer's /ai-review override first-principles ... product-shape on #26384. This backport ships the same default on the 0.11.1 line; no RFC records it on release/0.11.0.

PR picked/skipped why
#26384 picked one doc conflict in docs/system-specs/modules/cli.md, took only this PR's sentence on the kas install row (cli.sh, install.sh, cli_kas.py, install guides)
#26445 picked clean (website/electron/gateway-supervisor.js)
#26463 picked, adapted the five sites in modules 0.11.0 has (dashboard/handlers/crew_log.py, knowledge/artifact_ingest.py, remote_session/worker.py, subagent_manager/admission/gate.py, subagent_manager/terminal.py); shared_crew/admission_entry.py, sprites_session/terminal.py, the dropped-task guard test that lists them (test_dashboard_task_refs.py), test_runtime_background_task_retention.py and test_shared_crew_admission_failures.py are not on 0.11.0, so only the crew-log flush hunk has a regression test here
#26597 picked P0; fix(kas): refuse a file write a fallbackAction deny spec does not cover. Doc and comment conflicts only
#26456 skipped fix(security): match the chmod/chown rows on argv. Needs main-only modules and a main-only change to the same gate

rc.1 test failures: fixed by #26729 (merged; it backports #25703 and #24718's cron DST label fix), so this PR does not carry them.

Version and CHANGELOG:

Backwards compatibility

Compatible: every pick is a fix already shipped on main; the installers keep accepting install.sh --kas and add --no-kas to skip the engine.

Removes nothing: no key, route, record or public function is removed.

Tests

Manual verification

N/A - each fix is a cherry-pick of a main PR that carries its own evidence; the tests above run on this branch.

Related Issues

Part of the 0.11.1 hotfix (#26487, #26594, #26627, #26648, #26536, #26149, #26729). Release PRs take issue-gate: waived.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title
  • 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

Contributed under the project's Apache-2.0 license.

@bolichen97
bolichen97 requested review from a team and KiroCrewCaptain as code owners October 31, 2026 13:10
@bolichen97
bolichen97 requested review from CrysisDeu and removed request for a team October 31, 2026 13:10
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 31, 2026
@github-actions

github-actions Bot commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ✅ PASS

UX-level review of c85447b1c5fc056c7516a4f7ba533ba89461b102 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

UX-Verdict: PASS

Nothing to check. The diff adds and changes no dashboard UI. The Electron change is a quit-time spawn guard with no rendered output, and the rest is backend fixes, docs, version bumps and installer terminal text.

[UX-REVIEWED] c85447b

Description read

[DESCRIPTION-READ] 7a27a50094522642e5594a72f7f8157a9938db2874779637bd757813beedcae2

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 .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 Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of c85447b1c5fc056c7516a4f7ba533ba89461b102 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: the four picks are fixes already merged on main, the KAS-by-default installer change carries the owner's recorded decision, and a failed engine download only warns, so installs don't break.

Notes

[DESIGN-REVIEWED] c85447b

Description read

[DESCRIPTION-READ] 98e064a13e914d5ef611fdb306f61b2a646c092b5d40e97a72e4c612804eee97

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 Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — 🔴 BLOCK

Premise-level review of c85447b1c5fc056c7516a4f7ba533ba89461b102 — 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

product-shape change without accepted RFC: the CLI installers now download the KAS engine by default. The only record is an override on main for #26384, and nothing is recorded for this head.

What this change ships

Inventory (8 items) — 7 justified

Intent: ship four fixes already merged on main to 0.11.1 users as hotfix RC 0.11.1-rc.2. This is a FIX (a backport).

  1. install.sh/cli.sh download the pinned KAS engine by default; --no-kas skips it — product shape, no recorded decision
  2. A failed KAS download only warns and prints the retry command — justified
  3. Quitting the desktop app during startup no longer leaves a gateway running — justified
  4. Crew-log flush and four other background tasks are held until they finish instead of being dropped — justified
  5. A KAS spec's fallbackAction: "deny" write rule is now enforced — justified
  6. kirocrew kas install CLI and doc rows match the new default — justified
  7. Version files bumped to 0.11.1-rc.2 — justified
  8. The in-flight [0.11.1] CHANGELOG section lists the picked fixes and the filelock bump — justified

Blockers

product-shape change without accepted RFC: installers download KAS by default — product shape
Trigger: item 1. In install.sh, -WITH_KAS=0 becomes +WITH_KAS=1. In cli.sh, +INSTALL_KAS="${KIROCREW_INSTALL_KAS:-1}" is added. The base RFC list has no entry covering an installer KAS default; grepping kas|install finds only rfc-agent-plugins-load-installed.md: draft. The description itself says "no RFC records it on release/0.11.0". The override it cites belongs to #26384's head, not to c85447b1.
Outcome: every one-line install on the 0.11.1 line gets an extra engine download nobody chose, with no decision recorded for this branch's head.
Clears when: an RFC covering the installer's default KAS download is merged non-draft on the base (then push or rebase), or a maintainer posts /ai-review override first-principles c85447b1c5fc056c7516a4f7ba533ba89461b102 product-shape: <decision>.

[FIRST-PRINCIPLES-REVIEWED] c85447b

Description read

[DESCRIPTION-READ] 9d24c7518f04370281828f35cd565388a1c57db609f85dc0e91204e2e0c0b2ee

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.

@github-actions

github-actions Bot commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed c85447b1c5fc056c7516a4f7ba533ba89461b102 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] c85447b

Verdict parsed from the review's SHA-scoped output markers for commit c85447b1c5fc056c7516a4f7ba533ba89461b102.

False positive or not applicable? A repository writer can comment:
/ai-review override fable c85447b1c5fc056c7516a4f7ba533ba89461b102 false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — 🔴 changes requested (blocking)

GPT 6.1 found at least one blocking issue that must be resolved before merging c85447b1c5fc056c7516a4f7ba533ba89461b102. 1 of 1 blocking finding(s) are security-class and were withheld from adjudication, so the blocking verdict stands.

This comment is updated in place on each push.

BLOCKING -- src/kiro_crew/platform/allowed_paths.py:143 -- Literal workspace names become write-allowlist wildcards

pattern = posixpath.join(cwd, pattern)
Trigger: A POSIX KAS session in /work/project? sets fallbackAction: "deny" and allowedPaths: ["out"]. record_session_projection() → anchor_write_allowlist() → anchor() produces /work/project?/out; _write_target_uncovered() accepts /work/project2/out/report.txt.
Outcome: An AUTO_APPROVE turn approves a sibling-directory write outside the relative restriction. Both lexical and resolved target spellings match the broadened glob, so symlink verification does not refuse it.
Anchor: residual/security
Fix: Keep substituted cwd/home and resolved-directory prefixes literal during matching, or fail closed when those prefixes contain wildcard characters.
[BLOCK-MERGE] c85447b
[GPT-REVIEWED] c85447b

Adjudication (Opus 5.5) — is blocking on each finding proportionate?

F1 (fenced): the defect is real, but it needs the session's own directory, or the home directory, to have ? or * in its name. Operators almost never name directories that way.

  • What it needs: fallbackAction: "deny" is set, which produces the allowlist (kas_write_allowlist.py:31-34). A relative entry is joined to the cwd as raw text, with no escaping (allowed_paths.py:143, and the ~/home join at :139). parse() refuses []{}\$ but allows ? and * (allowed_paths.py:68,110), and this check applies only to the entry, not to the cwd. A sibling directory that matches the widened glob also has to exist. No retry or fallback undoes the approval: once the target is covered, _tier_spec_write_allowlist returns None (gate_tiers.py:1279-1282), so the write moves on to the later rows, and under AUTO_APPROVE it runs.

  • What it costs: what leaks is the spec author's own self-restriction, not the POLICY ∩ PROFILE ceiling, which still runs at its own rows. The module says it mirrors kiro-cli's own raw join of the entry onto the working directory (allowed_paths.py:1-7), so the native harness would read the same path the same way. The fix is small: escape the cwd/home prefix, or refuse when it holds a wildcard.

    [ADJUDICATION] c85447b total=0 uphold=0 downgrade=0
    [GPT-ADJUDICATED] c85447b

    [ADJUDICATION-FENCED] c85447b fenced=1 flagged=1
    FLAG F1 src/kiro_crew/platform/allowed_paths.py:143 -- Reaching it needs the session cwd or home directory itself to contain a literal ? or *, a sibling directory that matches, and AUTO_APPROVE; the widening affects only the spec's own fallbackAction allowlist, not the POLICY ∩ PROFILE ceiling, and matches kiro-cli's own raw join of the entry onto the cwd (allowed_paths.py:1-7).
    [GPT-ADJUDICATED-FENCED] c85447b

🏷️ Fenced finding(s) machine-flagged as likely edge case

The security fence keeps these findings blocking regardless of adjudication; the only clearance path is a human override. The annotate-only pass judged the condition combination each one requires extreme, and pre-drafted the override rationale below. A repository writer must independently verify a rationale before posting it — it is machine-authored, and a wrong override on a security-class finding ships exactly the class the fence exists to stop.

  • F1 src/kiro_crew/platform/allowed_paths.py:143 — Reaching it needs the session cwd or home directory itself to contain a literal ? or *, a sibling directory that matches, and AUTO_APPROVE; the widening affects only the spec's own fallbackAction allowlist, not the POLICY ∩ PROFILE ceiling, and matches kiro-cli's own raw join of the entry onto the cwd (allowed_paths.py:1-7).

    /ai-review override gpt c85447b1c5fc056c7516a4f7ba533ba89461b102 <class>: Reaching it needs the session cwd or home directory itself to contain a literal ? or *, a sibling directory that matches, and AUTO_APPROVE; the widening affects only the spec's own fallbackAction allowlist, not the POLICY ∩ PROFILE ceiling, and matches kiro-cli's own raw join of the entry onto the cwd (allowed_paths.py:1-7).
    

Replace <class> with the class that records what you did. These findings are real, so for an accepted risk use owner-decision (link the decision comment in the reason) or deferred #<N> (a tracked follow-up issue); false-positive only if you verified the finding is wrong. Not not-applicable: that class files the override as a rule mismatch in the audit.

False positive or not applicable? A repository writer can comment:
/ai-review override gpt c85447b1c5fc056c7516a4f7ba533ba89461b102 false-positive: <reason, 20 to 500 characters>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 31, 2026
@CrysisDeu

Copy link
Copy Markdown
Contributor

The v0.11.1-rc.1 Release run (41201776274) also failed Test Release Candidate SHA (5):

  • test/test_cron.py::TestFormatSchedule::test_cron_expr_with_timezone ('10:00 PM' in 'At 11:00 PM PST, Monday through Friday')
  • test_single_digit_hour_with_timezone
  • test_config_timezone_fallback

This is the DST label bug #24717. It is fixed on main by #24718 (3faf71c133, schedule.py plus test/test_cron_timezone_display.py), and release/0.11.0 doesn't have that fix. Until about 2026-11-01T09:00Z the failure is deterministic, so an rc.2 cut from this PR as it stands will fail shard 5 again. git cherry-pick -x 3faf71c133 applies cleanly on top of release/0.11.0. It was the second commit in #26729.

Backport to release/0.11.0, one commit:

- fix(install): download the pinned kas engine in cli installers (#26384)
  (cherry picked from commit 0a49fdb); docs/system-specs/modules/cli.md
  took only this PR's sentence on the `kas install` row.
- fix(electron): never spawn a gateway once the app is quitting (#26445)
  (cherry picked from commit 6a4d4e0), clean.
- fix: hold the remaining dropped asyncio tasks until they finish (#26463)
  (cherry picked from commit 8b3c698); adapted: the five sites in modules
  0.11.0 has (crew_log, artifact_ingest, remote_session/worker,
  subagent_manager admission gate and terminal). shared_crew/admission_entry.py,
  sprites_session/terminal.py and their tests are not on 0.11.0, so neither is
  the dropped-task guard test that lists them. The crew-log publisher test
  gains the held-task assertions, with `_flush` stubbed as on main.
- fix(kas): refuse a file write a fallbackAction deny spec does not cover
  (#26597) (cherry picked from commit 84c2aaa); doc and comment
  conflicts only: kept 0.11.0's wording around the sub-agent rules (#25361
  is not on 0.11.0) and dropped main's "v2 parity" acp-client block for the
  same reason.

Not picked: #26456 needs main-only `security/argv_rules.py` and the s3
download narrowing of #24562 in `deny_gate.py`. The v0.11.1-rc.1 Release
run's test failures are fixed by #26729.

All three version files move from 0.11.1-rc.1 to 0.11.1-rc.2, and the
in-flight [0.11.1] CHANGELOG section gains the four fixes and the
filelock bump of #26536 and the cron label fix of #26729 (#24718), both
already on the branch. #26149 and #25703 are test-only and have no
reader-visible line.
@bolichen97 bolichen97 changed the title chore(release): cut 0.11.1-rc.2 with three more hotfix fixes chore(release): cut 0.11.1-rc.2 with four more hotfix fixes Oct 31, 2026
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 31, 2026
@github-actions

Copy link
Copy Markdown
Contributor

Sensitive Change Review

Sensitive files changed:

  • src/kiro_crew/hook_runtime/gate_tiers.py

  • src/kiro_crew/hook_runtime/tool_identity.py

  • src/kiro_crew/hooks.py

  • No reasoning template for c85447b 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: c85447b1c5fc056c7516a4f7ba533ba89461b102
- 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:

@bolichen97

Copy link
Copy Markdown
Contributor Author

/ai-review override first-principles c85447b product-shape: Owner decision (bolichen, on-call): 0.11 runs KAS by default, so installers ship the pinned KAS engine. Same decision as #26384 on main; backported unchanged for 0.11.1.

@github-actions

Copy link
Copy Markdown
Contributor

AI-review override not recorded: product-shape is not a release class. Use /ai-review override <fable|gpt|design|ux|first-principles|scope|all> <current-sha> <class>[ #N]: <reason, 20 to 500 characters>, where <class> is one of false-positive, not-applicable, pre-existing, evidence-supplied, infra-failure, reissue, deferred, owner-decision. What each class needs: docs/ci/ci-and-reviews.md#risk-release-records.

@bolichen97
bolichen97 merged commit a27d362 into release/0.11.0 Oct 31, 2026
30 of 41 checks passed
@bolichen97
bolichen97 deleted the backport/0.11.1-rc2 branch October 31, 2026 13:53
@github-actions github-actions Bot removed the readiness: checking Automated validation is still running label Oct 31, 2026
@github-actions

Copy link
Copy Markdown
Contributor

Security Scope Review -- ✅ PASS

Which legitimate operations does c85447b1c5fc056c7516a4f7ba533ba89461b102 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): 13 classified, 0 newly refused, 13 still allowed.
  • ubuntu-latest (platform posix): 13 classified, 0 newly refused, 13 still allowed.
  • windows-latest (platform windows): 9 classified, 0 newly refused, 9 still allowed.

No candidate flipped from allowed to refused.

Scope-Verdict: PASS

Reason: The new restriction is a gate tier for file-write tool calls. It is not part of the shell deny composite (SHELL_DENY_TIERS still lists only sensitive-bash, exfil and deny-rules), so no shell candidate can return a verdict on it. That gap is named below. A second new refusal, which rejects remote-workspace agents that carry a shell allowlist, is also unadjudicated. No regression has been confirmed by the script, so neither gap blocks.

TIGHTENING: New deny_policy gate tier spec-write-allowlist (GateTier("spec-write-allowlist", "deny_policy", _tier_spec_write_allowlist)), plus a remote-workspace refusal (if _restricts_replaced_builtins(agent.get("permissions")) or _carries_spec_allowlist(agent):) -- the one threat it refuses: on KAS, a file write by a spec that sets toolsSettings.fs_write.fallbackAction: "deny" to a path its allowedPaths do not cover ("refuses a file write ... unless every target edit_target_candidates names is covered both as written ... and in every symlink-resolved spelling").

ADJUDICATED: 18 candidates, 0 confirmed newly-refused

PLATFORM COVERAGE: posix rows cover the installer flows this diff changes (bash install.sh --no-kas, --voice, KIROCREW_INSTALL_KAS=0 sh cli.sh), shell writes in the project directory, and ~ paths under ~/.kiro. Those probe the write-target and ~-expansion assumptions on any shell tier. windows rows cover a drive-letter path with \ separators (Set-Content C:\Users\...), entries spelled with $env: and with %VAR% (the spellings allowed_paths.parse now refuses), kirocrew.exe kas install, and a .venv\Scripts\python.exe test run. Those probe the ON_WINDOWS branch assumptions (backslash folding, ASCII case folding, drive root). any rows cover the platform-neutral kirocrew kas install/status/--force and config set agent.kas_bundle. I wrote no UNC-path candidate as a shell row. The UNC refusal applies only to allowedPaths entries, which only the write tier reads.

UNADJUDICATED:

  • Guard: _tier_spec_write_allowlist / _write_target_uncovered (src/kiro_crew/hook_runtime/gate_tiers.py), matched by platform/allowed_paths.py. Call path: the spec's toolsSettings.fs_write|write.fallbackAction: "deny" → kas_permissions.spec_write_allowlist → to_client_custom_agent writes _crewWriteAllowlist → KasHarness sets projected_write_allowlist(custom_agents, active_agent, handle._bound_cwd) (re-read via switched_write_allowlist on a mode switch) → _build_permission_event stamps spec_write_allowlist and spec_write_cwd → ToolCall.from_event → the spec-write-allowlist tier, which runs on is_edit_call / fs_write calls only. It also reaches the kirocrew chat gate through cli_chat._answer_permission and side turns through acp_turns._to_llm_event. Legitimate inputs I would have submitted:
    • an fs_write to relative src/app.py with allowedPaths: ["src/**"] and a bound cwd;
    • the same write when _bound_cwd is empty. The tier refuses it ("relative path and the session directory is unknown"), and it is worth confirming that every KAS session start binds a cwd;
    • on macOS, an fs_write to /tmp/out.txt with allowedPaths: ["/tmp/**"], where /tmp resolves to /private/tmp. This exercises the entry-prefix resolution;
    • a write under a home directory reached through a symlink, with allowedPaths: ["~/proj"];
    • on Windows, an fs_write to C:\Users\dev\proj\src\x.py with allowedPaths: ["src\\**"], and with ["C:\\Users\\Dev\\proj\\**"] (mixed case);
    • on Windows, allowedPaths: ["%USERPROFILE%\\proj\\**"] and ["$env:USERPROFILE\\proj\\**"]. Both cover nothing by design, so every write that spec makes is refused on KAS, where kiro-cli might expand them. This is a narrowing I can argue but not confirm;
    • delete_file on a covered path;
    • a diff-only edit whose diff_path is relative. The tier refuses it as unanchored even when it would sit under an allowed entry.
  • Guard: _carries_spec_allowlist in _remote_workspace_agents (src/kiro_crew/acp/runtime_remote_workspace.py). Call path: a remote-workspace session start → _remote_workspace_agents → AcpRuntimeError(REMOTE_AGENT_RESTRICTED) for any projected entry holding _crewShellAllowlist or _crewWriteAllowlist. This newly refuses an agent whose spec sets only shell.denyByDefault, with an allowedCommands list and no file rules, when it starts in a remote workspace. Before this diff that agent started there. Legitimate input I would have submitted: a remote-workspace session for a spec with toolsSettings.shell: {denyByDefault: true, allowedCommands: ["git status"]}. The guard is stated as intentional (the allowlist cannot bind MCP-mapped built-ins), and I cannot confirm it with a verdict.

[SCOPE-REVIEWED] c85447b

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.

2 participants