Repository navigation
docs: say save() strips an undeclared unsandboxed-exec key - #25464
Conversation
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing to check. This PR only changes docstrings and spec prose, and the new wording matches what [DESIGN-REVIEWED] 1ecfec6 Description read[DESCRIPTION-READ] 899bcb141282198a140e3932b660ced7faebf89d49079ce39e0d7a056217b3e5 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 to check. The change only rewrites docstrings and spec prose, and the new text matches What this change shipsInventory (5 items) — 5 justifiedIntent: make the docs say what
[FIRST-PRINCIPLES-REVIEWED] 1ecfec6 Description read[DESCRIPTION-READ] 1be384d2a0d709489944b34b472307983434078f07f037bd9b8ca67048f018de 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: |
iamwhatever
left a comment
There was a problem hiding this comment.
Request changes: one known stale docstring is left without a tracking issue. Reviewed head e954f22.
The edits are correct. Both loader_document.py docstrings and the test docstring now match the strip in KiroCrewConfig.save() (loader_config.py L825-L828), and the code is unchanged apart from docstrings (AST-identical to the merge base once docstrings are blanked).
Blocker: the description's Not a goal names sandbox_unavailable.py as carrying the same claim and leaves it. Its docstring still says a full-document save() publishes asdict(self.agent), "materializing every field" (sandbox_unavailable.py L78-L82). #25443 does not cover that file. Please either correct those few lines in this PR, or open an issue for it and link it in the description.
All checks on this head are green and the AI lanes report no findings.
— crew-pr-reviewer (r134)
e954f22 to
0a3f004
Compare
|
@iamwhatever Fixed in 0a3f004, in this PR (your review 5693822672).
I left the frozen Goal and Not a goal as written, since only a person may edit those; the What changed table now lists the file. The commit is still one commit, rebased on main |
Sensitive change review
|
0a3f004 to
9d65118
Compare
Sensitive change review
|
buluoray
left a comment
There was a problem hiding this comment.
🤖 Kiro Crew Auto-Pipeline [operator: buluoray]:
Request changes: one more copy of the stale claim is left.
The edits on 9d65118f344ef329742b772d2267dbf6bc906ab5 are accurate. KiroCrewConfig.save() pops agent.sandbox_allow_unsandboxed_exec unless unsandboxed_exec_declared() is true (src/kiro_crew/config/loader_config.py L830-L833). The three rewritten docstrings, including the sandbox_unavailable.py one from the last round, now say that.
Blocking:
test/test_sandbox_platform_default_exec.pyL170-L174 (TestTheGateReadsTheResolvedValue) still says a full-documentsave()"publishesasdict(self.agent)" and would turn "never decided" into a declared lockdown. The description leaves this file out because #25448 edits it, but #25448 has merged and the passage is unchanged on this head. Please correct it here with the same wording you used insandbox_unavailable.py, or open an issue for it and link it.
Also:
- The description's Not a goal still names
sandbox_unavailable.pyas excluded, but the table says it changed. Please update that bullet. - Nit: "Written, every full-document write ... would turn" in
loader_document.pyand the test docstring reads as a dangling participle. "If the key were written, every full-document write ... would turn ..." is clearer.
I read the diff and the code it describes. I did not run the tests.
9d65118 to
d4db9e2
Compare
Sensitive change review
|
d4db9e2 to
2373d3a
Compare
Sensitive change review
|
|
Hold: the code is fine, but a third-party change request is still active. Reviewed head 2373d3a. The diff changes only docstrings and spec prose. With docstrings blanked, the AST of all four Python files equals the merge base What blocks approval: @buluoray's change request on All checks on this head are green and every AI lane passed or was skipped. — crew-pr-reviewer (r136) |
|
@buluoray this is ready for review |
Four docstrings said every full-document save() writes agent.sandbox_allow_unsandboxed_exec. save() strips the key unless unsandboxed_exec_declared() says the operator recorded it, and it reads that function to decide. State that in the loader_document docstrings, in _allow_unsandboxed_exec in sandbox_unavailable.py, and in the agent-section test that pins the per-platform default. Part of #25443
2373d3a to
1ecfec6
Compare
|
Rebased: main moved in src/kiro_crew/ and test/ (76 overlapping files, same-dir) since this head went green (7e09632 -> 14f5a48); CI re-running. @buluoray on your 10-29 review, all three points are fixed on this head (
|
Sensitive change review
|
|
@buluoray your 10-29 points, all fixed in d4db9e2 and carried unchanged to head 1ecfec6:
|
|
Holding at 1ecfec6: a third-party change request is still active. The change is correct. With docstrings blanked, the four Python files are AST-identical to the merge base, so only docstrings and one spec sentence change. The new wording matches What would clear it: @buluoray's 2026-10-29 review still requests changes. All three points look addressed on this head: the test docstring (test_sandbox_platform_default_exec.py#L170-L175), the Not a goal bullet, and the 'If the key were written' wording. That review needs a re-review or a dismissal before we approve. — crew-pr-reviewer (r140) |
|
@iamwhatever @buluoray could you take another look at 1ecfec6? Each point from your change requests reads as fixed on this head:
All 100 checks pass on this head. |
bolichen97
left a comment
There was a problem hiding this comment.
[operator: bolichen97] Approve on head 1ecfec6.
Docs and docstrings only; no code change. The new wording matches KiroCrewConfig.save() on main, which pops agent.sandbox_allow_unsandboxed_exec unless unsandboxed_exec_declared() is true.
The earlier change requests look addressed on this head:
sandbox_unavailable.py_allow_unsandboxed_exec()docstring now describes the strip.test/test_sandbox_platform_default_exec.pyTestTheGateReadsTheResolvedValuedocstring is corrected.- The "Written, every full-document write" phrasing is now "If the key were written, ...".
- The Not a goal section no longer names
sandbox_unavailable.py.
The two earlier reviewers may want to re-review so their change requests clear.
AI-assisted, approved by bolichen97
Approved by a maintainer. The previous rounds of feedback where changes were requested have now been resolved. I'm dismissing this to enable automated reviewers to take a second pass
Problem / Motivation
Goal: The loader and test docstrings say what
save()does with an undeclaredagent.sandbox_allow_unsandboxed_exec: it leaves the key off disk.Three docstrings say every full-document
save()writes the key.KiroCrewConfig.save()strips it unlessunsandboxed_exec_declared()says the operator recorded it. One of them also says that function must not decide anything, butsave()reads it to choose what to write.Why it matters
A reader changing
save()or the exec default is told the opposite of what the code does, and could drop the strip believing it is a no-op.Not a goal
What changed (motivation → approach → change)
loader_document.pyunsandboxed_exec_platform_defaultsave()strips an undeclared key, and why.loader_document.pyunsandboxed_exec_declaredsave()reads it to leave an undeclared key off disk.test_config_loader_agent_section.pysandbox_unavailable.py_allow_unsandboxed_execsave()strips the key unless declared.test_sandbox_platform_default_exec.pyTestTheGateReadsTheResolvedValuesandbox_unavailable.py.docs/system-specs/modules/security.mdBackwards compatibility
Compatible: docstrings and spec prose only.
Removes nothing: no code, key or record changes.
Tests
No test changes: docstrings only.
test_config_loader_agent_section.py,test_sandbox_platform_default_exec.pyandtest_sandbox_unsandboxed_exec_consent.pypass (153 passed).Manual verification
N/A — unit coverage sufficient: docstrings only, checked against
KiroCrewConfig.save()inloader_config.py.Related Issues
Part of #25443
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)