Repository navigation
fix(auto-improvement): pin utf-8 on the agent_runner subprocess call - #27104
Conversation
Pin UTF-8 on the text-mode subprocess call(s) in src/kiro_crew/apps/builtins/auto_improvement/spine/agent_runner.py and drop the file's row from .github/subprocess-encoding-baseline.txt, so the subprocess-encoding gate stops tolerating it.
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing to check: no screen changes. The diff only pins the encoding on a backend subprocess call and removes a CI baseline row, so no screenshot is needed. [UX-REVIEWED] 6ce7b8a Description read[DESCRIPTION-READ] 62b6dc4348e76456e9a0b57915e628fd5d5bb657ce10ee0679d932117c0cc930 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) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: this fixes the root cause, a decode that depended on the locale, using the same pattern as the earlier burn-down PRs. Notes
[DESIGN-REVIEWED] 6ce7b8a Description read[DESCRIPTION-READ] 81b163d70d46f1484cba8a256e804e088950c8b4cad9914aa6f66c8ae880c898 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 blocks: the one-line encoding pin removes a decode defect with known provenance (#5249), and the baseline row is removed under a documented one-way ratchet. What this change shipsInventory (2 items) — 2 justifiedIntent: make the auto-improvement agent's streamed output decode as UTF-8 on every platform, Windows included. This is a FIX.
The website test and act-baseline hunks in the patch belong to #27040, which is already on main (215a485), so the author did not change them. [FIRST-PRINCIPLES-REVIEWED] 6ce7b8a Description read[DESCRIPTION-READ] eaa4c8912930e272cc137cb727e0e59a175f2c6612a3d9d0781891d9560486e1 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: |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Tier 1 auto-approve: fix (2 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, security checklist all-NO, no reviewer-lane findings, readiness signal fresh for this head SHA, declared category matches the diff. Category: adds encoding="utf-8" to one text-mode Popen in the auto-improvement agent runner, so its output decodes the same on every host.
Problem / Motivation
Goal: A text-mode subprocess call in 1 file decodes its child's output as UTF-8 on every platform instead of the locale's code page.
The subprocess-encoding gate counts, per file, the sites the repo still tolerates. This PR retires:
1 src/kiro_crew/apps/builtins/auto_improvement/spine/agent_runner.py.Why it matters
Without an explicit encoding,
subprocess.run(..., text=True)decodes with the locale code page, which on Windows is not UTF-8: non-ASCII output becomes mojibake or raisesUnicodeDecodeError(#5249). The repo lists every such call per file in.github/subprocess-encoding-baseline.txtand the gate lets each count only shrink.Not a goal
What changed (motivation → approach → change)
Each call below now passes
encoding="utf-8"beside itstext=True, and the file's baseline line is removed.src/kiro_crew/apps/builtins/auto_improvement/spine/agent_runner.py:316: thepopen_limited(...)call in_spawn_sandboxed_agent.github/subprocess-encoding-baseline.txt: removed1 src/kiro_crew/apps/builtins/auto_improvement/spine/agent_runner.py.Backwards compatibility
Compatible: the call's return type and the output's meaning are unchanged; only the decode encoding is pinned.
Removes nothing: a baseline entry is a tolerance the gate stops needing, not a capability.
Tests
RATCHET_SCOPE_WHOLE_TREE=1 python3 scripts/check_subprocess_encoding.pyexits 1:::error file=src/kiro_crew/apps/builtins/auto_improvement/spine/agent_runner.py::text-mode subprocess call without ....python3 scripts/check_subprocess_encoding.py --test && python3 scripts/check_subprocess_encoding.pyexit 0.test/test_ai_agent_discovery_helpers.py: 49 passed with the edit (importskiro_crew.apps.builtins.auto_improvement.spine.agent_runner)test/test_kas_cli_floor.py: 77 passed with the edit (importskiro_crew.apps.builtins.auto_improvement.spine.agent_runner)Manual verification
N/A: the subprocess-encoding gate checks the changed call directly.
Sibling sweep
Swept:
python3 scripts/check_subprocess_encoding.py, the gate's own scan of every text-mode subprocess call. The other files it reports keep their rows in.github/subprocess-encoding-baseline.txt; each is retired by its own PR.Sibling:
src/kiro_crew/apps/builtins/auto_improvement/spine/agent_runner.py:316-- changed herePattern harvest
Rule candidate: none new; the rule is already enforced by
scripts/check_subprocess_encoding.py(a baselined CI gate that flags every text-mode subprocess call without an explicit encoding), and this PR retires baseline entries under it.Screenshots / video
Why no screenshot: nothing renders; this change pins a subprocess decode encoding and a CI baseline only.
No issue exists; burn-down of the subprocess-encoding gate baseline (precedents #20146, #20142, #20147).