⚡ Bolt: iter_json_objects JSON 디코딩 성능 최적화

.gitignore

@@ -0,0 +1,4 @@
+__pycache__/
+*.py[cod]
+.coverage
+.pytest_cache/

.jules/bolt.md

@@ -1,3 +1,6 @@
 ## 2024-06-21 - Python JSON Decoding Optimization
 **Learning:** In Python, string slicing `text[index:]` inside a loop can cause O(N^2) complexity and severe memory copying overhead. When decoding JSON incrementally from a large text blob, `json.JSONDecoder().raw_decode(text, index)` can parse from a given index without slicing. Combining this with `text.find("{", index)` to skip irrelevant characters is significantly faster than `enumerate(text)`.
 **Action:** Always prefer `raw_decode(text, index)` and `string.find()` over string slicing and character-by-character iteration when scanning large files for JSON objects.
+## 2024-06-23 - `iter_json_objects` 최적화
+**Learning:** Python의 `json.JSONDecoder().raw_decode()`를 사용할 때 문자열을 하나씩 순회하며 슬라이싱(`text[index:]`)을 수행하면, O(N^2)의 메모리 할당 및 복사 작업이 발생하여 매우 큰 병목(Bottleneck)이 될 수 있습니다.
+**Action:** `str.find("{", index)`를 사용하여 JSON 객체의 시작 위치를 빠르게 건너뛰고, `raw_decode(text, index)`에서 제공하는 `idx` 인자를 활용해 슬라이싱 없이 직접 파싱을 수행하여 최적화합니다.

PR_GOVERNANCE_AUDIT.md

@@ -15,6 +15,7 @@ OpenCode decides; GitHub Actions mutates.
 - OpenCode app-token merges are deprecated; keep app tokens for review publication, not mechanical branch mutation.
 - OpenCode approval publication must be bounded. Peer GitHub Checks can be awaited, but the approval step itself must time out instead of running for hours; the current central limit is a 45 minute approval step with 81 peer-check probes at 30 seconds.
 - Tool failures are not source findings. Model failure, API transient, update-branch `422/403`, fork/write-permission failure, conflict, failed checks, and stale review state must be reported as distinct scheduler outcomes.
+- Developer experience and user experience are separate review surfaces. Reviews must adopt helpful sibling-repo automation, review, setup, documentation, and product-flow patterns when they reduce friction, and flag noisy automation, false failures, misleading status, repeated waiting, or URL-only diagnostics as experience defects instead of treating them as neutral implementation detail.
 
 ## Live Repository Inventory
 
@@ -59,6 +60,27 @@ Live generated: 2026-06-23 04:18 KST. PR #28 post-merge refresh: 2026-06-23 16:0
 | `codec-carver` | Recent PR #94 was merged by `app/opencode-agent`, and the repo still has legacy `Scheduled PR Review Merge`. | Native auto-merge path for current-head approved PRs. | OpenCode app as merge actor. |
 | `VibeSec` | PR #108 had native auto-merge enabled; #106 merged by `app/github-actions`; #109 merged by human. | Keep native auto-merge as preferred waiting path. | Repo-by-repo actor inconsistency. |
 
+## DX/UX Transfer Decisions
+
+Developer experience means the maintainer, reviewer, CI operator, and future
+contributor experience. User experience means the product user, documentation
+reader, PR reader, and status-check reader experience. PR review must evaluate
+both separately; a change can improve one while harming the other.
+
+| Repo | Borrow because it helps DX/UX | Improve because it creates friction | Central action |
+|---|---|---|---|
+| `.github` | Same-head manual evidence and `--match-head-commit` make self-modifying workflow changes reviewable without pretending stale base-branch checks are current. | Stale `pull_request_target` failures, long polling review runs, and cancelled helper checks can become misleading review noise. | Serialize Strix before OpenCode, bound approval runtime, and require failed-check explanations instead of URL-only comments. |
+| `naruon` | Strict required checks, stale review dismissal, and changed-file Mermaid flow DAGs make review evidence easier to audit. | `BEHIND` or outdated approvals can look merge-ready unless the head SHA is treated as the review boundary. | Re-review every updated head and require an exact changed-file evidence path plus a Change Flow DAG before approval. |
+| `pg-erd-cloud` | GitHub Actions bot merges with head guards give a clear mechanical actor for merges. | Repo-local autofix workflows are useful there, but centralizing autofix would widen mutation scope too far. | Keep GitHub Actions as the merge actor and leave autofix workflows repo-local. |
+| `VibeSec` | Native auto-merge examples show a lower-friction waiting path after current-head approval. | Mixed human, OpenCode, and GitHub Actions merge actors make audit trails harder to interpret. | Prefer native auto-merge or GitHub Actions mutation; do not let OpenCode merge directly. |
+| `bandscope` | Broad required checks encode repo-specific release, build, SBOM, and security expectations. | A central script would be noisy if it tried to reinterpret every required check itself. | Let GitHub native auto-merge and rulesets interpret required checks. |
+| `newsdom-api` | Required quality gates and security checks give API changes stronger release evidence. | Central review comments that only point at failing check URLs do not help an API maintainer fix the failure. | Require failed-check root cause, source location when available, fix direction, and rerun command. |
+| `scopeweave` | Strix self-test and the central scheduler are useful rollout fixtures. | Claiming the rollout complete without a representative current-head trace would be premature. | Keep it on the central path and require a live trace before declaring update/merge behavior proven. |
+| `clearfolio` | Direct guarded merge works while auto-merge is intentionally off. | Treating it like an auto-merge repo would create confusing expectations. | Use immediate guarded merge only after same-head evidence, unresolved-thread check, and head guard pass. |
+| `codec-carver` | Existing native merge behavior can be retained once current-head evidence is clean. | Legacy OpenCode app-token merge creates a second mechanical actor and weakens audit consistency. | Replace legacy scheduled merge with the central GitHub Actions scheduler. |
+| `ContextualWisdomLab.github.io` | Site and documentation changes make reader-facing UX review concrete. | Review comments that say only that a check failed do not help the site reader or maintainer understand the issue. | Treat documentation clarity, homepage behavior, and status-check explanations as UX surfaces. |
+| `contextual-orchestrator` | No central pattern is present yet, so it can be onboarded deliberately instead of accidentally. | Silent unmanaged status is easy to miss in organization-level governance. | Either opt it into the central workflows or explicitly mark it unmanaged. |
+
 ## Current Scheduler Contract
 
 The checked-in scheduler already does the minimal central path:
@@ -131,5 +153,6 @@ PR #36: block: merge conflict: DIRTY
 - `clearfolio` PR #13 and `codec-carver` PR #98 were opened as thin rollouts. `clearfolio` PR #13 is now merged at `4bc17c6`; `codec-carver` PR #98 remains the thin rollout that deletes the legacy OpenCode app-token merge workflow.
 - `clearfolio` PR #13 first failed Strix run `28027843973` because `opencode.jsonc` was missing. Later current-head proof used manual Strix run `28051319530` and manual OpenCode run `28051665082`; the final approval named the changed review-tooling files and head `5fe1791d48ddcf03dbc365cc6fa407e7cbe70a89` before guarded merge.
 - `.github` PR #42 exposed that central approval normalization should not accept generic path-looking evidence when exact current-head changed files are available. The OpenCode workflow now writes `git diff --name-only --find-renames "$PR_MERGE_BASE" "$PR_HEAD_SHA"` to `OPENCODE_CHANGED_FILES_FILE`, gives the isolated review workspace `changed-files.txt`, and the normalizer rejects `APPROVE` unless the approval names one of those exact files.
+- `.github` PR #42 same-head OpenCode run `28070438305` exposed a second decode gap: model output reading tolerated invalid UTF-8, but approval-summary repair still read `OPENCODE_APPROVAL_REPAIR_EVIDENCE_FILE` as strict UTF-8. DeepSeek produced a repairable control block, then normalization failed on byte `0xea` in bounded evidence. Evidence repair now reads lossy UTF-8 so a damaged transcript byte cannot prevent source-backed normalization.
 - `codec-carver` PR #98 already has base `opencode.jsonc`. PR #98 now pins the central scheduler instead of downloading from `main`; same-head Strix run `28030439830` and OpenCode runs `28030438605`/`28030439065` were still in progress at the 2026-06-23 22:48 KST snapshot.
 - `.github` PR #38 exposed two central gaps after PR #37 merged: the `review_dispatch` reason lost the `same-head Strix and OpenCode dispatched` contract string, and `failed_status_checks()` treated failed PR-target Strix check runs as blockers even when a later manual `strix` status could supersede them. Commit `7be2d99` restores the reason string, materializes PR-head scheduler policy as non-executed data for Strix self-test, and ignores stale Strix check-run failures when the same head has a successful `strix` status context. Manual Strix run `28030448032` had passed self-test and was still running `Run Strix (quick)` at the 2026-06-23 22:48 KST snapshot.

README.md

@@ -7,6 +7,9 @@ The public GitHub organization profile lives in [profile/README.md](profile/READ
 Homepage: https://contextualwisdomlab.github.io/
 
 PR governance live audit: [PR_GOVERNANCE_AUDIT.md](PR_GOVERNANCE_AUDIT.md).
+The audit includes repository-by-repository DX/UX transfer decisions: what the
+central workflow borrows because it reduces friction, and what it rejects
+because it adds noise or misleading review experience.
 
 ## PR review and merge policy
 
@@ -27,9 +30,12 @@ privileged `pull_request_target` OpenCode workflow.
 OpenCode approval is evidence-gated. Before approval, the review summary must
 name changed files, CodeGraph or structural MCP evidence, a Change Flow DAG,
 100% test coverage evidence, 100% docstring coverage evidence, and a concrete
-PoC/execution result. The PoC can be a temporary scratch repro, focused test,
-lint, security check, performance probe, or UI verification command, but it must
-be actually run and cited. Scratch PoC files are not committed.
+PoC/execution result. It must also split `Developer experience:` from
+`User experience:` so maintainability/review/CI friction is not confused with
+product, documentation, review-comment, or status-check reader outcomes. The PoC
+can be a temporary scratch repro, focused test, lint, security check,
+performance probe, or UI verification command, but it must be actually run and
+cited. Scratch PoC files are not committed.
 
 Failed GitHub Checks are not reviewed as URL lists. OpenCode must explain the
 failed check name, failing step, source-backed file and line when available,
@@ -51,3 +57,7 @@ Operational cases folded into the central policy:
 - `naruon#745`: new OpenCode review-flow work improves Mermaid output by
   replacing generic risk sketches with changed-file flow DAGs. The central
   workflow carries that review contract while keeping the self-test drift fix.
+- Cross-repo DX/UX: helpful sibling-repo patterns should be adopted when they
+  reduce maintainer, reviewer, CI-operator, contributor, user, or reader
+  friction. Noisy automation, repeated waiting, false failures, misleading
+  statuses, and URL-only diagnostics are treated as review-experience defects.

scripts/ci/opencode_review_normalize_output.py

@@ -97,7 +97,8 @@
     "compatibility/convention:",
     "breaking-change/backcompat:",
     "performance:",
-    "design/ux:",
+    "developer experience:",
+    "user experience:",
     "security/privacy:",
 )
 
@@ -223,6 +224,14 @@ def approval_repair_evidence_file() -> Path | None:
     return None
 
 
+def read_text_lossy(path: Path) -> str | None:
+    """Read text while preserving progress across invalid UTF-8 bytes."""
+    try:
+        return path.read_text(encoding="utf-8", errors="replace")
+    except OSError:
+        return None
+
+
 def section_between_markers(text: str, marker: str) -> str:
     """Return a markdown section body from a bounded evidence file."""
     marker_line = f"## {marker}"
@@ -296,7 +305,8 @@ def build_approval_repair_summary(summary: str, evidence_text: str) -> str | Non
 Compatibility/convention: changed workflow/script conventions and compatibility surfaces were checked in bounded evidence.
 Breaking-change/backcompat: deployment evidence and changed-file history were checked for backward-compatibility risk.
 Performance: changed surfaces were checked for performance risk in bounded evidence.
-Design/UX: changed files did not identify a UI-facing design surface; bounded evidence was reviewed.
+Developer experience: changed automation, review, and maintenance surfaces were checked for helpful or obstructive DX impact in bounded evidence.
+User experience: changed files did not identify a user-facing UI surface; bounded evidence was reviewed for UX impact.
 Security/privacy: workflow-token, review-gate, and repository-automation security/privacy boundaries were checked in bounded evidence.
 """
     return f"{summary.rstrip()}\n{repair}"
@@ -312,9 +322,8 @@ def repair_approval_summary(reason: str, summary: str) -> str:
     evidence_file = approval_repair_evidence_file()
     if evidence_file is None:
         return summary
-    try:
-        evidence_text = evidence_file.read_text(encoding="utf-8")
-    except OSError:
+    evidence_text = read_text_lossy(evidence_file)
+    if evidence_text is None:
         return summary
 
     repaired_summary = build_approval_repair_summary(summary, evidence_text)
@@ -452,14 +461,23 @@ def iter_json_objects(text: str) -> list[Any]:
         # OpenCode exports may contain prose around the JSON control object.
         pass
 
-    for index, character in enumerate(text):
-        if character != "{":
+    index = 0
+    while True:
+        index = text.find("{", index)
+        if index == -1:
+            break
+        next_index = index + 1
+        while next_index < len(text) and text[next_index] in " \t\r\n":
+            next_index += 1
+        if next_index < len(text) and text[next_index] not in {'"', "}"}:
+            index += 1
             continue
         try:
-            value, _ = decoder.raw_decode(text[index:])
+            value, _ = decoder.raw_decode(text, index)
+            values.append(value)
         except json.JSONDecodeError:
-            continue
-        values.append(value)
+            pass
+        index += 1
 
     return values
 
@@ -481,7 +499,7 @@ def main(argv: list[str]) -> int:
     expected_head_sha, expected_run_id, expected_run_attempt, output_file_arg = argv[1:]
     output_file = Path(output_file_arg)
     try:
-        output_text = output_file.read_text(encoding="utf-8")
+        output_text = output_file.read_text(encoding="utf-8", errors="replace")
     except OSError as exc:
         print(f"cannot read OpenCode output file: {exc}", file=sys.stderr)
         return 65