From 6ddae875c6de3a0d80f49781f643f9ab00f21069 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Thu, 30 Jul 2026 01:38:28 +0000
Subject: [PATCH 1/5] fix(review): defer R coverage-evidence to peer R CMD
 check when package cannot load

The OpenCode approval gate blocks every R package with compiled Imports
(aFIPC, kaefa, nonnest2). run_r_test_coverage installs distribution packages
only (r-base + covr + testthat, no build tools) and runs
testthat::test_dir("tests/testthat") without installing the package-under-test
or its Imports. For a package that hard-Imports a compiled dependency such as
mirt, every test that touches pkg:: then fails at loadNamespace() (observed:
'make: not found', 'FAIL 7' on aFIPC #189 run 30505299347), so coverage-evidence
reports failure and OpenCode posts REQUEST_CHANGES even though the repository's
own required R CMD check job (which installs dependencies) runs the full suite
and passes.

This block already defers to the required peer R CMD check evidence when testthat
itself is unavailable. Extend that same deferral to the case where the package
under test cannot be loaded (missing compiled Imports or the package is not
installed in the distribution-only sandbox). The deferral is safe: the peer
R CMD check is a required gate that installs dependencies and runs the full
suite, so a genuine R test failure still blocks the merge; no gate is weakened.
When the package does load, tests run exactly as before and real failures are
still caught.

Validated: full contract suite green (674 passed), incl. the workflow
shell-syntax and opencode contract gates.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AzUnTqFnQqhRbaopvDdag7
---
 .../workflows/opencode-review-dispatch.yml    | 125 ++++++++++++++-
 scripts/ci/r_coverage_peer_gate.py            | 120 ++++++++++++++
 scripts/ci/test_strix_quick_gate.sh           |   6 +-
 tests/test_opencode_agent_contract.py         |  35 ++++-
 tests/test_r_coverage_peer_gate.py            | 147 ++++++++++++++++++
 5 files changed, 425 insertions(+), 8 deletions(-)
 create mode 100644 scripts/ci/r_coverage_peer_gate.py
 create mode 100644 tests/test_r_coverage_peer_gate.py

diff --git a/.github/workflows/opencode-review-dispatch.yml b/.github/workflows/opencode-review-dispatch.yml
index 94f5cbb7c..359deb29a 100644
--- a/.github/workflows/opencode-review-dispatch.yml
+++ b/.github/workflows/opencode-review-dispatch.yml
@@ -730,6 +730,7 @@ jobs:
           summary_file="${RUNNER_TEMP}/coverage-evidence.md"
           summary_output_file="${RUNNER_TEMP}/coverage-evidence-output.md"
           failures=0
+          r_peer_check_required=0
 
           append() {
             printf '%s\n' "$*" >>"$summary_file"
@@ -803,6 +804,61 @@ jobs:
             rm -f "$log_file"
           }
 
+          run_r_package_testthat() {
+            local package_name="$1"
+            local log_file rc classification
+            log_file="$(mktemp)"
+            append "### R package testthat suite"
+            append ""
+            append '```text'
+            append_command \
+              Rscript -e 'lib <- Sys.getenv("R_LIBS_USER"); .libPaths(c(lib, .libPaths())); testthat::test_dir("tests/testthat")'
+            set +e
+            timeout --kill-after=20 900 setpriv \
+              --reuid "$OPENCODE_SANDBOX_UID" \
+              --regid "$OPENCODE_SANDBOX_GID" \
+              --clear-groups \
+              env \
+              -u ACTIONS_ID_TOKEN_REQUEST_TOKEN \
+              -u ACTIONS_ID_TOKEN_REQUEST_URL \
+              -u ACTIONS_RUNTIME_TOKEN \
+              -u GH_TOKEN \
+              -u GITHUB_TOKEN \
+              GITHUB_ENV=/dev/null \
+              GITHUB_PATH=/dev/null \
+              GITHUB_OUTPUT=/dev/null \
+              GITHUB_STEP_SUMMARY=/dev/null \
+              BASH_ENV=/dev/null \
+              UV_NO_BUILD=1 \
+              HOME=/work/.opencode-sandbox-home \
+              XDG_CACHE_HOME=/work/.opencode-sandbox-cache \
+              PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin" \
+              Rscript -e 'lib <- Sys.getenv("R_LIBS_USER"); .libPaths(c(lib, .libPaths())); testthat::test_dir("tests/testthat")' \
+              >"$log_file" 2>&1
+            rc=$?
+            set -e
+            emit_captured_log "$log_file"
+            append '```'
+            append ""
+            if [ "$rc" -eq 0 ]; then
+              append "- Result: PASS"
+            elif classification="$(
+              python3 -I "$GITHUB_WORKSPACE/scripts/ci/r_coverage_peer_gate.py" \
+                classify-testthat \
+                --log "$log_file" \
+                --package "$package_name" 2>/dev/null
+            )"; then
+              append "- Result: PASS"
+              append "- Reason: ${classification}; direct sandbox failures are deferred only to a successful current-head peer R CMD check."
+              r_peer_check_required=1
+            else
+              append "- Result: FAIL (exit ${rc})"
+              failures=$((failures + 1))
+            fi
+            append ""
+            rm -f "$log_file"
+          }
+
           run_and_capture_advisory() {
             local label="$1"
             shift
@@ -1326,6 +1382,7 @@ jobs:
           }
 
           run_r_test_coverage() {
+            local package_name
             ensure_r_runtime
             if ! command -v Rscript >/dev/null 2>&1; then
               append "### R test coverage"
@@ -1344,8 +1401,10 @@ jobs:
               Rscript -e 'required <- c("covr", "testthat"); missing <- required[!vapply(required, requireNamespace, logical(1), quietly = TRUE)]; if (length(missing)) stop("signed distribution coverage packages unavailable: ", paste(missing, collapse = ", "))'
             if [ -f DESCRIPTION ]; then
               if [ -d tests/testthat ]; then
-                run_and_capture "R package testthat suite" \
-                  Rscript -e 'lib <- Sys.getenv("R_LIBS_USER"); .libPaths(c(lib, .libPaths())); if (!requireNamespace("testthat", quietly = TRUE)) { message("testthat unavailable in coverage runner; deferring to required peer R CMD check evidence."); quit(status = 0) }; testthat::test_dir("tests/testthat")'
+                package_name="$(
+                  Rscript -e 'pkg <- tryCatch(read.dcf("DESCRIPTION")[1, "Package"], error = function(e) ""); cat(pkg)'
+                )"
+                run_r_package_testthat "$package_name"
               else
                 append "### R package testthat suite"
                 append ""
@@ -1706,6 +1765,9 @@ jobs:
             else
               append "- Test evidence: supported repository test suites passed"
               append "- Docstring evidence: configured repository docstring gates passed or docstring coverage was advisory"
+              if [ "$r_peer_check_required" -eq 1 ]; then
+                append "- R test evidence: deferred package-load failures require a successful current-head peer R CMD check"
+              fi
             fi
           else
             append "- Result: FAIL"
@@ -6527,6 +6589,40 @@ jobs:
             return 2
           }
 
+          coverage_defers_to_r_cmd_check() {
+            printf '%s\n' "${COVERAGE_EVIDENCE_SUMMARY:-}" |
+              grep -Fq -- "- R test evidence: deferred package-load failures require a successful current-head peer R CMD check"
+          }
+
+          collect_successful_r_cmd_check_evidence() {
+            local output_file="$1"
+            if ! gh pr checks "$PR_NUMBER" \
+              --repo "$GH_REPOSITORY" \
+              --json name,state,workflow >"$output_file"; then
+              return 1
+            fi
+            python3 "$GITHUB_WORKSPACE/scripts/ci/r_coverage_peer_gate.py" \
+              require-check \
+              --checks-json "$output_file" >/dev/null
+          }
+
+          require_r_cmd_check_for_deferred_coverage() {
+            local checks_file
+            if ! coverage_defers_to_r_cmd_check; then
+              return 0
+            fi
+            checks_file="$(mktemp)"
+            if collect_github_checks_with_retry \
+              collect_successful_r_cmd_check_evidence "$checks_file"; then
+              rm -f "$checks_file"
+              printf 'Verified successful current-head R CMD check after bounded R coverage deferral.\n'
+              return 0
+            fi
+            rm -f "$checks_file"
+            printf '::notice::R package-load coverage deferral cannot authorize approval without a successful current-head R CMD check.\n'
+            return 1
+          }
+
           stop_without_review_after_model_unavailable() {
             local body
             body="$(printf '%s\n' \
@@ -6643,6 +6739,10 @@ jobs:
               return 0
             fi
 
+            if ! require_r_cmd_check_for_deferred_coverage; then
+              return 1
+            fi
+
             failed_check_evidence_file="$(mktemp)"
             if ! collect_open_code_scanning_alerts "$failed_check_evidence_file"; then
               printf '::notice::Current-head model-unavailable evidence fallback skipped because open code-scanning alerts could not be read.\n'
@@ -7027,6 +7127,27 @@ jobs:
                   stop_failed_check_fallback_unavailable
                 fi
               fi
+              if ! require_r_cmd_check_for_deferred_coverage; then
+                body="$(printf '%s\n' \
+                  "## Pull request overview" \
+                  "" \
+                  "OpenCode reviewed the current-head source evidence but R package tests were deferred after bounded package-load-only failures." \
+                  "" \
+                  "## Approval hold" \
+                  "" \
+                  "### Successful current-head R CMD check evidence is required" \
+                  "- Problem: coverage-evidence deferred package-load-only testthat failures, but no successful current-head R CMD check was found." \
+                  "- Root cause: deterministic coverage deferral is safe only when the repository's peer R CMD check installs dependencies and passes on this exact head." \
+                  "- Fix: add or repair the R CMD check workflow and rerun it successfully on the current head." \
+                  "- Regression test: Keep deferred R coverage fail-closed unless a successful R CMD check is present." \
+                  "" \
+                  "- Result: WAITING_FOR_R_CMD_CHECK" \
+                  "- Head SHA: \`${HEAD_SHA}\`" \
+                  "- Workflow run: ${RUN_ID}" \
+                  "- Workflow attempt: ${RUN_ATTEMPT}"
+                )"
+                hold_approval_without_review "WAITING_FOR_R_CMD_CHECK" "$body"
+              fi
               unresolved_reviewer_threads_file="$(mktemp)"
               reviewer_thread_review_body_file="$(mktemp)"
               if ! collect_unresolved_reviewer_threads "$unresolved_reviewer_threads_file"; then
diff --git a/scripts/ci/r_coverage_peer_gate.py b/scripts/ci/r_coverage_peer_gate.py
new file mode 100644
index 000000000..af19b02bc
--- /dev/null
+++ b/scripts/ci/r_coverage_peer_gate.py
@@ -0,0 +1,120 @@
+#!/usr/bin/env python3
+"""Gate R coverage deferral on bounded testthat and peer-check evidence."""
+
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+import re
+import sys
+from typing import Any, Sequence
+
+
+MAX_LOG_BYTES = 2_000_000
+PACKAGE_NAME_RE = re.compile(r"[A-Za-z][A-Za-z0-9.]*\Z")
+FAIL_SUMMARY_RE = re.compile(r"^\[\s*FAIL\s+(\d+)\s*\|", re.MULTILINE)
+ERROR_BLOCK_RE = re.compile(r"^Error \(", re.MULTILINE)
+PACKAGE_NOT_FOUND_CONDITION_RE = re.compile(
+    r"^<packageNotFoundError/error/condition>$",
+    re.MULTILINE,
+)
+MISSING_PACKAGE_RE = re.compile(r"there is no package called ['\"]([^'\"]+)['\"]")
+R_CMD_CHECK_RE = re.compile(r"\br[\s_-]*cmd[\s_-]*check\b", re.IGNORECASE)
+
+
+def classify_testthat_failure(text: str, package: str) -> bool:
+    """Return whether every testthat failure is the uninstalled package under test."""
+    if not PACKAGE_NAME_RE.fullmatch(package):
+        return False
+    summaries = FAIL_SUMMARY_RE.findall(text)
+    if not summaries or "Error: Test failures" not in text:
+        return False
+    failure_count = int(summaries[-1])
+    if failure_count <= 0:
+        return False
+    error_count = len(ERROR_BLOCK_RE.findall(text))
+    condition_count = len(PACKAGE_NOT_FOUND_CONDITION_RE.findall(text))
+    missing_packages = MISSING_PACKAGE_RE.findall(text)
+    return (
+        error_count == failure_count
+        and condition_count == failure_count
+        and len(missing_packages) == failure_count
+        and all(name == package for name in missing_packages)
+    )
+
+
+def has_successful_r_cmd_check(checks: Any) -> bool:
+    """Return whether check JSON contains a successful R CMD check workflow."""
+    if not isinstance(checks, list):
+        return False
+    for check in checks:
+        if not isinstance(check, dict):
+            continue
+        if str(check.get("state") or "").upper() != "SUCCESS":
+            continue
+        label = f"{check.get('workflow') or ''} {check.get('name') or ''}"
+        if R_CMD_CHECK_RE.search(label):
+            return True
+    return False
+
+
+def _read_bounded_text(path: Path) -> str | None:
+    """Read a regular bounded log, returning None for unsafe or unreadable input."""
+    try:
+        if not path.is_file() or path.is_symlink() or path.stat().st_size > MAX_LOG_BYTES:
+            return None
+        return path.read_text(encoding="utf-8", errors="replace")
+    except OSError:
+        return None
+
+
+def _read_json(path: Path) -> Any:
+    """Read JSON from a regular bounded file, returning None on invalid input."""
+    text = _read_bounded_text(path)
+    if text is None:
+        return None
+    try:
+        return json.loads(text)
+    except json.JSONDecodeError:
+        return None
+
+
+def parse_args(argv: Sequence[str] | None = None) -> argparse.Namespace:
+    """Parse the requested peer-gate operation."""
+    parser = argparse.ArgumentParser()
+    subparsers = parser.add_subparsers(dest="command", required=True)
+
+    classify = subparsers.add_parser("classify-testthat")
+    classify.add_argument("--log", type=Path, required=True)
+    classify.add_argument("--package", required=True)
+
+    require_check = subparsers.add_parser("require-check")
+    require_check.add_argument("--checks-json", type=Path, required=True)
+    return parser.parse_args(argv)
+
+
+def main(argv: Sequence[str] | None = None) -> int:
+    """Run the selected R coverage peer-evidence gate."""
+    args = parse_args(argv)
+    if args.command == "classify-testthat":
+        text = _read_bounded_text(args.log)
+        if text is not None and classify_testthat_failure(text, args.package):
+            print(
+                "testthat failures were exclusively packageNotFoundError "
+                f"conditions for package {args.package}"
+            )
+            return 0
+        print("testthat failure is not safely deferrable", file=sys.stderr)
+        return 1
+
+    checks = _read_json(args.checks_json)
+    if has_successful_r_cmd_check(checks):
+        print("successful current-head R CMD check evidence found")
+        return 0
+    print("successful current-head R CMD check evidence was not found", file=sys.stderr)
+    return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/scripts/ci/test_strix_quick_gate.sh b/scripts/ci/test_strix_quick_gate.sh
index 497999e91..a60ab448b 100755
--- a/scripts/ci/test_strix_quick_gate.sh
+++ b/scripts/ci/test_strix_quick_gate.sh
@@ -547,7 +547,11 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
 	assert_file_contains "$workflow_file" "r-cran-covr" "opencode R coverage uses the signed distribution covr package instead of mutable CRAN resolution"
 	assert_file_contains "$workflow_file" "r-cran-testthat" "opencode R coverage uses the signed distribution testthat package instead of mutable CRAN resolution"
 	assert_file_contains "$workflow_file" "R package testthat suite" "opencode R package coverage requires package testthat evidence"
-	assert_file_contains "$workflow_file" "testthat unavailable in coverage runner; deferring to required peer R CMD check evidence." "opencode R package tests defer only when testthat cannot be installed in the coverage runner"
+	assert_file_contains "$workflow_file" "r_coverage_peer_gate.py" "opencode R package coverage classifies bounded package-load-only failures with trusted code"
+	assert_file_contains "$workflow_file" "- R test evidence: deferred package-load failures require a successful current-head peer R CMD check" "opencode R package coverage records explicit peer-check deferral evidence"
+	assert_file_contains "$workflow_file" "require_r_cmd_check_for_deferred_coverage" "opencode approval verifies deferred R evidence against current-head peer checks"
+	assert_file_contains "$workflow_file" "WAITING_FOR_R_CMD_CHECK" "opencode approval fails closed when deferred R coverage lacks successful peer evidence"
+	assert_file_not_contains "$workflow_file" 'if (!is.na(pkg) && !requireNamespace(pkg, quietly = TRUE))' "opencode R coverage does not skip the entire test suite merely because the source package is not preinstalled"
 	assert_file_contains "$workflow_file" "covr package_coverage unavailable after package tests; treating missing-line report as advisory." "opencode R package coverage does not block on covr installation reproduction after tests pass"
 	assert_file_contains "$workflow_file" "signed distribution coverage packages unavailable" "opencode R coverage verifies distribution-provided covr/testthat are loadable"
 	assert_file_contains "$workflow_file" "repository: ContextualWisdomLab/.github" "opencode required workflow checks out the central source repository"
diff --git a/tests/test_opencode_agent_contract.py b/tests/test_opencode_agent_contract.py
index 0e791907f..f4bc62f3a 100644
--- a/tests/test_opencode_agent_contract.py
+++ b/tests/test_opencode_agent_contract.py
@@ -1760,11 +1760,11 @@ def test_opencode_privileged_review_security_boundaries_are_fail_closed():
     assert syntax_step < measure_step
     assert "\n      - name:" not in measure.split("\n        run: |", 1)[1]
     assert 'UV_NO_BUILD: "1"' in measure
-    assert measure.count("GITHUB_ENV=/dev/null") == 2
-    assert measure.count("GITHUB_PATH=/dev/null") == 2
-    assert measure.count("GITHUB_OUTPUT=/dev/null") == 2
-    assert measure.count("GITHUB_STEP_SUMMARY=/dev/null") == 2
-    assert measure.count("BASH_ENV=/dev/null") == 2
+    assert measure.count("GITHUB_ENV=/dev/null") == 3
+    assert measure.count("GITHUB_PATH=/dev/null") == 3
+    assert measure.count("GITHUB_OUTPUT=/dev/null") == 3
+    assert measure.count("GITHUB_STEP_SUMMARY=/dev/null") == 3
+    assert measure.count("BASH_ENV=/dev/null") == 3
     assert "uv sync --project" not in measure
     assert "uv run --no-project" not in measure
     assert "uv run --no-build" not in measure
@@ -2316,3 +2316,28 @@ def test_slow_peer_wait_matches_only_image_validation_checks():
             or re.search(package_build_pattern, candidate, re.IGNORECASE) is not None
         )
         assert slow_build_match is slow_build_expected, candidate
+
+
+def test_r_package_load_deferral_requires_current_head_r_cmd_check():
+    """R package-load-only failures may defer only to explicit peer evidence."""
+    workflow = Path(".github/workflows/opencode-review-dispatch.yml").read_text(
+        encoding="utf-8"
+    )
+    marker = (
+        "- R test evidence: deferred package-load failures require a successful "
+        "current-head peer R CMD check"
+    )
+
+    assert "run_r_package_testthat" in workflow
+    assert "r_coverage_peer_gate.py" in workflow
+    assert marker in workflow
+    assert "require_r_cmd_check_for_deferred_coverage" in workflow
+    assert workflow.count("require_r_cmd_check_for_deferred_coverage") == 3
+    assert "WAITING_FOR_R_CMD_CHECK" in workflow
+    assert "testthat unavailable in coverage runner" not in workflow
+    assert (
+        "pkg <- tryCatch(read.dcf(\"DESCRIPTION\")[1, \"Package\"]" in workflow
+    )
+    assert (
+        "if (!is.na(pkg) && !requireNamespace(pkg, quietly = TRUE))" not in workflow
+    )
diff --git a/tests/test_r_coverage_peer_gate.py b/tests/test_r_coverage_peer_gate.py
new file mode 100644
index 000000000..e0be03e5f
--- /dev/null
+++ b/tests/test_r_coverage_peer_gate.py
@@ -0,0 +1,147 @@
+"""Tests for fail-closed R coverage deferral and peer-check evidence."""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+import runpy
+import sys
+
+import pytest
+
+from scripts.ci import r_coverage_peer_gate as gate
+
+
+def test_classifies_only_own_package_not_found_failures() -> None:
+    """Every reported failure must be packageNotFoundError for the package under test."""
+    text = """\
+Error ('test-one.R:1:1'): first
+<packageNotFoundError/error/condition>
+Error in `loadNamespace(x)`: there is no package called 'aFIPC'
+Error ('test-two.R:2:1'): second
+<packageNotFoundError/error/condition>
+Error in `loadNamespace(x)`: there is no package called 'aFIPC'
+[ FAIL 2 | WARN 0 | SKIP 1 | PASS 3 ]
+Error: Test failures
+"""
+    assert gate.classify_testthat_failure(text, "aFIPC")
+
+
+def test_rejects_invalid_or_mixed_test_failures() -> None:
+    """Malformed names, assertions, other packages, and count drift fail closed."""
+    assertion = "[ FAIL 1 | WARN 0 | SKIP 0 | PASS 0 ]\nError: Test failures\n"
+    other_package = """\
+Error ('test-one.R:1:1'): first
+<packageNotFoundError/error/condition>
+Error in `loadNamespace(x)`: there is no package called 'mirt'
+[ FAIL 1 | WARN 0 | SKIP 0 | PASS 0 ]
+Error: Test failures
+"""
+    mismatched = other_package.replace("FAIL 1", "FAIL 2").replace("mirt", "aFIPC")
+    zero_failures = "[ FAIL 0 | WARN 0 | SKIP 0 | PASS 1 ]\nError: Test failures\n"
+
+    assert not gate.classify_testthat_failure("", "aFIPC")
+    assert not gate.classify_testthat_failure(
+        "[ FAIL 1 | WARN 0 | SKIP 0 | PASS 0 ]", "aFIPC"
+    )
+    assert not gate.classify_testthat_failure(zero_failures, "aFIPC")
+    assert not gate.classify_testthat_failure(assertion, "aFIPC")
+    assert not gate.classify_testthat_failure(other_package, "aFIPC")
+    assert not gate.classify_testthat_failure(mismatched, "aFIPC")
+    assert not gate.classify_testthat_failure(other_package, "../aFIPC")
+
+
+def test_requires_successful_r_cmd_check_workflow() -> None:
+    """Only a successful check whose workflow/name identifies R CMD check qualifies."""
+    checks = [
+        {"workflow": "R CMD check", "name": "check", "state": "SUCCESS"},
+        {"workflow": "Other", "name": "test", "state": "FAILURE"},
+    ]
+    assert gate.has_successful_r_cmd_check(checks)
+    assert gate.has_successful_r_cmd_check(
+        [{"workflow": "", "name": "R-CMD-check", "state": "success"}]
+    )
+    assert not gate.has_successful_r_cmd_check(
+        [{"workflow": "R CMD check", "name": "check", "state": "FAILURE"}]
+    )
+    assert not gate.has_successful_r_cmd_check(
+        [{"workflow": "CI", "name": "tests", "state": "SUCCESS"}]
+    )
+    assert not gate.has_successful_r_cmd_check({"checks": checks})
+    assert not gate.has_successful_r_cmd_check(["invalid"])
+
+
+def test_cli_classifies_log_and_check_json(tmp_path: Path, capsys) -> None:
+    """Both CLI modes accept bounded valid evidence and reject invalid JSON."""
+    log = tmp_path / "testthat.log"
+    log.write_text(
+        "Error ('x.R:1:1'): x\n"
+        "<packageNotFoundError/error/condition>\n"
+        "Error in `loadNamespace(x)`: there is no package called 'pkg'\n"
+        "[ FAIL 1 | WARN 0 | SKIP 0 | PASS 0 ]\n"
+        "Error: Test failures\n",
+        encoding="utf-8",
+    )
+    checks = tmp_path / "checks.json"
+    checks.write_text(
+        json.dumps([{"workflow": "R CMD check", "name": "check", "state": "SUCCESS"}]),
+        encoding="utf-8",
+    )
+
+    assert gate.main(["classify-testthat", "--log", str(log), "--package", "pkg"]) == 0
+    assert gate.main(["require-check", "--checks-json", str(checks)]) == 0
+
+    checks.write_text("{", encoding="utf-8")
+    assert gate.main(["require-check", "--checks-json", str(checks)]) == 1
+    assert (
+        gate.main(
+            [
+                "require-check",
+                "--checks-json",
+                str(tmp_path / "missing-checks.json"),
+            ]
+        )
+        == 1
+    )
+    assert "not found" in capsys.readouterr().err
+
+
+def test_cli_rejects_unsafe_or_oversized_logs(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """Missing, symlinked, and oversized artifacts never authorize deferral."""
+    missing = tmp_path / "missing.log"
+    assert (
+        gate.main(["classify-testthat", "--log", str(missing), "--package", "pkg"])
+        == 1
+    )
+
+    target = tmp_path / "target.log"
+    target.write_text("small", encoding="utf-8")
+    link = tmp_path / "link.log"
+    link.symlink_to(target)
+    assert gate.main(["classify-testthat", "--log", str(link), "--package", "pkg"]) == 1
+
+    target.write_bytes(b"x" * (gate.MAX_LOG_BYTES + 1))
+    assert gate.main(["classify-testthat", "--log", str(target), "--package", "pkg"]) == 1
+
+    monkeypatch.setattr(Path, "is_file", lambda _path: (_ for _ in ()).throw(OSError()))
+    assert gate._read_bounded_text(target) is None
+
+
+def test_script_entrypoint_returns_cli_status(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """The executable entrypoint propagates the fail-closed CLI status."""
+    missing = tmp_path / "missing.log"
+    script = Path("scripts/ci/r_coverage_peer_gate.py")
+    monkeypatch.setattr(
+        sys,
+        "argv",
+        [str(script), "classify-testthat", "--log", str(missing), "--package", "pkg"],
+    )
+
+    with pytest.raises(SystemExit) as raised:
+        runpy.run_path(str(script), run_name="__main__")
+
+    assert raised.value.code == 1

From 7b471194be190dec21a591583e1b36383eacf935 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Thu, 30 Jul 2026 04:54:43 +0000
Subject: [PATCH 2/5] ci(opencode-review): expand free OpenCode Zen reviewer
 pool

Add five verified free ($0 input/output), tool_call-capable OpenCode Zen
models to the public-PR review pool so it degrades gracefully instead of
exhausting to a hard block when the primary free models are rate-limited:

  - big-pickle           (active; reasoning + structured_output; 200K/32K)
  - nemotron-3-super-free (204800/128K; sibling of the configured ultra)
  - minimax-m2.5-free    (204800/131072; text; coding-strong)
  - mimo-v2.5-free       (200K/32K; reasoning + tool_call)
  - grok-code            (256K/256K; last-resort fallback)

Each model's tool_call flag, cost ($0/$0), and context/output limits were
taken from the authoritative models.dev registry (providers/opencode). The
established active models (nemotron-3-ultra-free, deepseek-v4-flash-free,
north-mini-code-free) keep their leading order so review quality is
unchanged; the additions are appended purely as deeper fallbacks, active
first. The runner already skips any model the gateway reports as
unavailable, so listing a served-but-deprecated model is strictly additive.

Contract test (tests/test_opencode_agent_contract.py) updated to pin the
new candidate list and provider entries; full suite green (710 passed).

Co-Authored-By: Claude <noreply@anthropic.com>
---
 .../workflows/opencode-review-dispatch.yml    | 47 ++++++++++++++++++-
 tests/test_opencode_agent_contract.py         | 24 +++++++++-
 2 files changed, 68 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/opencode-review-dispatch.yml b/.github/workflows/opencode-review-dispatch.yml
index 359deb29a..ba5083c9c 100644
--- a/.github/workflows/opencode-review-dispatch.yml
+++ b/.github/workflows/opencode-review-dispatch.yml
@@ -3456,6 +3456,51 @@ jobs:
                       "context": 256000,
                       "output": 64000
                     }
+                  },
+                  "big-pickle": {
+                    "name": "Big Pickle",
+                    "tool_call": true,
+                    "reasoning": true,
+                    "limit": {
+                      "context": 200000,
+                      "output": 32000
+                    }
+                  },
+                  "nemotron-3-super-free": {
+                    "name": "Nemotron 3 Super Free",
+                    "tool_call": true,
+                    "reasoning": true,
+                    "limit": {
+                      "context": 204800,
+                      "output": 128000
+                    }
+                  },
+                  "minimax-m2.5-free": {
+                    "name": "MiniMax-M2.5 Free",
+                    "tool_call": true,
+                    "reasoning": true,
+                    "limit": {
+                      "context": 204800,
+                      "output": 131072
+                    }
+                  },
+                  "mimo-v2.5-free": {
+                    "name": "MiMo V2.5 Free",
+                    "tool_call": true,
+                    "reasoning": true,
+                    "limit": {
+                      "context": 200000,
+                      "output": 32000
+                    }
+                  },
+                  "grok-code": {
+                    "name": "Grok Code Fast 1 Free",
+                    "tool_call": true,
+                    "reasoning": true,
+                    "limit": {
+                      "context": 256000,
+                      "output": 256000
+                    }
                   }
                 }
               },
@@ -3788,7 +3833,7 @@ jobs:
           # cost-efficient tier, cheaper than the legacy gpt-5 it replaced
           # ($1/$6 vs $1.25/$10 per 1M tokens) so the org OpenAI budget
           # stretches further between top-ups.
-          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
+          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/big-pickle opencode-free/nemotron-3-super-free opencode-free/minimax-m2.5-free opencode-free/mimo-v2.5-free opencode-free/grok-code ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
           # One attempt per model, then fall through to the next model. Retrying
           # the SAME model 5x let a rate-limited/hung leader consume the whole
           # step, so the pool never reached a healthy fallback model.
diff --git a/tests/test_opencode_agent_contract.py b/tests/test_opencode_agent_contract.py
index f4bc62f3a..58304ed2a 100644
--- a/tests/test_opencode_agent_contract.py
+++ b/tests/test_opencode_agent_contract.py
@@ -94,7 +94,12 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "${{ needs.validate-pr-metadata.outputs.is_private == 'false' "
         "&& 'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
-        "opencode-free/north-mini-code-free ' || '' }}"
+        "opencode-free/north-mini-code-free "
+        "opencode-free/big-pickle "
+        "opencode-free/nemotron-3-super-free "
+        "opencode-free/minimax-m2.5-free "
+        "opencode-free/mimo-v2.5-free "
+        "opencode-free/grok-code ' || '' }}"
     )
     candidates_text = candidates_match.group(1)
     assert candidates_text.startswith(conditional_public_candidate)
@@ -102,6 +107,11 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "opencode-free/nemotron-3-ultra-free",
         "opencode-free/deepseek-v4-flash-free",
         "opencode-free/north-mini-code-free",
+        "opencode-free/big-pickle",
+        "opencode-free/nemotron-3-super-free",
+        "opencode-free/minimax-m2.5-free",
+        "opencode-free/mimo-v2.5-free",
+        "opencode-free/grok-code",
         *candidates_text.removeprefix(conditional_public_candidate).split(),
     ]
     candidate_pairs = [candidate.split("/", 1) for candidate in candidates]
@@ -122,6 +132,11 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         ["opencode-free", "nemotron-3-ultra-free"],
         ["opencode-free", "deepseek-v4-flash-free"],
         ["opencode-free", "north-mini-code-free"],
+        ["opencode-free", "big-pickle"],
+        ["opencode-free", "nemotron-3-super-free"],
+        ["opencode-free", "minimax-m2.5-free"],
+        ["opencode-free", "mimo-v2.5-free"],
+        ["opencode-free", "grok-code"],
         ["github-models", "deepseek/deepseek-v3-0324"],
         ["openai", "gpt-5.6-luna"],
         ["openrouter", "deepseek/deepseek-v3.2"],
@@ -1257,7 +1272,12 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
         "needs.validate-pr-metadata.outputs.is_private == 'false' && "
         "'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
-        "opencode-free/north-mini-code-free ' || ''"
+        "opencode-free/north-mini-code-free "
+        "opencode-free/big-pickle "
+        "opencode-free/nemotron-3-super-free "
+        "opencode-free/minimax-m2.5-free "
+        "opencode-free/mimo-v2.5-free "
+        "opencode-free/grok-code ' || ''"
     ) in workflow
     assert (
         "github-models/deepseek/deepseek-v3-0324 "

From 6a0f9f6a98ecc3a0443a3c379b7fe8662bf870cf Mon Sep 17 00:00:00 2001
From: Seongho Bae <me@seonghobae.me>
Date: Thu, 30 Jul 2026 14:04:55 +0900
Subject: [PATCH 3/5] fix(review): bound active Zen model fallback pool

---
 .../workflows/opencode-review-dispatch.yml    | 49 +++++++---------
 scripts/ci/test_strix_quick_gate.sh           |  2 +-
 tests/test_opencode_agent_contract.py         | 58 +++++++++++++------
 3 files changed, 61 insertions(+), 48 deletions(-)

diff --git a/.github/workflows/opencode-review-dispatch.yml b/.github/workflows/opencode-review-dispatch.yml
index ba5083c9c..91ecded91 100644
--- a/.github/workflows/opencode-review-dispatch.yml
+++ b/.github/workflows/opencode-review-dispatch.yml
@@ -3436,8 +3436,8 @@ jobs:
                     "name": "DeepSeek V4 Flash Free",
                     "tool_call": true,
                     "limit": {
-                      "context": 256000,
-                      "output": 64000
+                      "context": 200000,
+                      "output": 128000
                     }
                   },
                   "north-mini-code-free": {
@@ -3466,22 +3466,22 @@ jobs:
                       "output": 32000
                     }
                   },
-                  "nemotron-3-super-free": {
-                    "name": "Nemotron 3 Super Free",
+                  "laguna-s-2.1-free": {
+                    "name": "Laguna S 2.1 Free",
                     "tool_call": true,
                     "reasoning": true,
                     "limit": {
-                      "context": 204800,
-                      "output": 128000
+                      "context": 256000,
+                      "output": 32000
                     }
                   },
-                  "minimax-m2.5-free": {
-                    "name": "MiniMax-M2.5 Free",
+                  "ling-3.0-flash-free": {
+                    "name": "Ling-3.0-flash Free",
                     "tool_call": true,
                     "reasoning": true,
                     "limit": {
-                      "context": 204800,
-                      "output": 131072
+                      "context": 262144,
+                      "output": 32768
                     }
                   },
                   "mimo-v2.5-free": {
@@ -3492,15 +3492,6 @@ jobs:
                       "context": 200000,
                       "output": 32000
                     }
-                  },
-                  "grok-code": {
-                    "name": "Grok Code Fast 1 Free",
-                    "tool_call": true,
-                    "reasoning": true,
-                    "limit": {
-                      "context": 256000,
-                      "output": 256000
-                    }
                   }
                 }
               },
@@ -3818,8 +3809,8 @@ jobs:
           NPM_CONFIG_IGNORE_SCRIPTS: "true"
           NO_COLOR: "1"
           # High-sensitivity review candidates only. Public repositories first
-          # use OpenCode Zen's anonymous Nemotron 3 Ultra, DeepSeek V4 Flash,
-          # and North Mini Code free endpoints. Their trial/free-period data
+          # use OpenCode Zen's anonymous active, zero-cost model endpoints.
+          # Their trial/free-period data
           # may be logged, retained, or used for product/model improvement, so
           # private repositories never include them. The pool then falls
           # through to
@@ -3833,7 +3824,7 @@ jobs:
           # cost-efficient tier, cheaper than the legacy gpt-5 it replaced
           # ($1/$6 vs $1.25/$10 per 1M tokens) so the org OpenAI budget
           # stretches further between top-ups.
-          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/big-pickle opencode-free/nemotron-3-super-free opencode-free/minimax-m2.5-free opencode-free/mimo-v2.5-free opencode-free/grok-code ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
+          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/laguna-s-2.1-free opencode-free/ling-3.0-flash-free opencode-free/big-pickle opencode-free/mimo-v2.5-free ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
           # One attempt per model, then fall through to the next model. Retrying
           # the SAME model 5x let a rate-limited/hung leader consume the whole
           # step, so the pool never reached a healthy fallback model.
@@ -3846,11 +3837,11 @@ jobs:
           OPENCODE_EXPORT_TIMEOUT_SECONDS: "180"
           OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "11700"
           OPENCODE_POOL_STEP_TIMEOUT_SECONDS: "12000"
-          # Keep cycling through the high-sensitivity candidate catalog until
-          # the retry budget or step timeout is exhausted; a single invalid
-          # cycle can be all provider formatting noise rather than review
-          # evidence.
-          OPENCODE_POOL_MAX_CYCLES: "0"
+          # A second pass through the same provider catalog repeats the same
+          # quota/format failures and can occupy the required check for hours.
+          # Exhaust each distinct candidate once, then publish the bounded
+          # model-unavailable fallback with current-head evidence.
+          OPENCODE_POOL_MAX_CYCLES: "1"
           OPENCODE_DYNAMIC_REVIEW_CADENCE: "true"
           OPENCODE_SMALL_CHANGE_FILE_THRESHOLD: "3"
           OPENCODE_MEDIUM_CHANGE_FILE_THRESHOLD: "20"
@@ -3864,14 +3855,14 @@ jobs:
           OPENCODE_UNKNOWN_CHANGE_TOTAL_BUDGET_SECONDS: "11700"
           OPENCODE_DYNAMIC_RUN_TIMEOUT_CAP_SECONDS: "5400"
           OPENCODE_DYNAMIC_TOTAL_BUDGET_CAP_SECONDS: "11700"
-          OPENCODE_DYNAMIC_MAX_CYCLES_CAP: "0"
+          OPENCODE_DYNAMIC_MAX_CYCLES_CAP: "1"
           OPENCODE_FREE_RUN_TIMEOUT_SECONDS: "600"
           # This installation currently reports a 4k request-body limit for
           # GitHub Models GPT-5 endpoints even though the public catalog is
           # larger. Keep the exact runtime failure visible without spending a
           # full medium/large cadence slot after the long-context candidate.
           OPENCODE_GITHUB_GPT5_RUN_TIMEOUT_SECONDS: "45"
-          OPENCODE_DYNAMIC_MAX_CYCLES: "0"
+          OPENCODE_DYNAMIC_MAX_CYCLES: "1"
           CENTRAL_REVIEW_PROCESS_FALLBACK_ELIGIBLE: ${{ steps.central_review_process_fallback_scope.outputs.eligible || 'false' }}
           CENTRAL_REVIEW_PROCESS_FALLBACK_SCOPE_LABEL: ${{ steps.central_review_process_fallback_scope.outputs.scope_label || 'unsupported' }}
           OPENCODE_CENTRAL_REVIEW_PROCESS_FALLBACK_RUN_TIMEOUT_SECONDS: "5400"
diff --git a/scripts/ci/test_strix_quick_gate.sh b/scripts/ci/test_strix_quick_gate.sh
index a60ab448b..c336c20da 100755
--- a/scripts/ci/test_strix_quick_gate.sh
+++ b/scripts/ci/test_strix_quick_gate.sh
@@ -709,7 +709,7 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
 	assert_file_contains "$workflow_file" 'continue-on-error: true' "opencode approval gate still runs after model-pool failure to publish a reason"
 	assert_file_contains "$workflow_file" 'OPENCODE_RUN_TIMEOUT_SECONDS: "5400"' "opencode primary review preserves legitimate full-hour provider sessions"
 	assert_file_contains "$workflow_file" 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "11700"' "opencode model pool exits before the step timeout so the approval gate can publish a reason"
-	assert_file_contains "$workflow_file" 'OPENCODE_POOL_MAX_CYCLES: "0"' "opencode model pool keeps cycling until the bounded retry budget or step timeout is exhausted"
+	assert_file_contains "$workflow_file" 'OPENCODE_POOL_MAX_CYCLES: "1"' "opencode model pool exhausts each candidate only once before bounded fallback"
 	assert_file_not_contains "$workflow_file" 'opencode-exhausted-retry:' "opencode model exhaustion retries stay owned by the least-privilege central scheduler"
 	assert_file_not_contains "$workflow_file" 'RETRY_DISPATCH_TOKEN' "opencode does not retain a recursive write-token dispatch path"
 	assert_file_contains "$workflow_file" "needs.coverage-evidence.result == 'success'" "opencode model pool only runs after coverage evidence passed"
diff --git a/tests/test_opencode_agent_contract.py b/tests/test_opencode_agent_contract.py
index 58304ed2a..12ffcb5ef 100644
--- a/tests/test_opencode_agent_contract.py
+++ b/tests/test_opencode_agent_contract.py
@@ -95,11 +95,10 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "&& 'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
         "opencode-free/north-mini-code-free "
+        "opencode-free/laguna-s-2.1-free "
+        "opencode-free/ling-3.0-flash-free "
         "opencode-free/big-pickle "
-        "opencode-free/nemotron-3-super-free "
-        "opencode-free/minimax-m2.5-free "
-        "opencode-free/mimo-v2.5-free "
-        "opencode-free/grok-code ' || '' }}"
+        "opencode-free/mimo-v2.5-free ' || '' }}"
     )
     candidates_text = candidates_match.group(1)
     assert candidates_text.startswith(conditional_public_candidate)
@@ -107,11 +106,10 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "opencode-free/nemotron-3-ultra-free",
         "opencode-free/deepseek-v4-flash-free",
         "opencode-free/north-mini-code-free",
+        "opencode-free/laguna-s-2.1-free",
+        "opencode-free/ling-3.0-flash-free",
         "opencode-free/big-pickle",
-        "opencode-free/nemotron-3-super-free",
-        "opencode-free/minimax-m2.5-free",
         "opencode-free/mimo-v2.5-free",
-        "opencode-free/grok-code",
         *candidates_text.removeprefix(conditional_public_candidate).split(),
     ]
     candidate_pairs = [candidate.split("/", 1) for candidate in candidates]
@@ -132,11 +130,10 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         ["opencode-free", "nemotron-3-ultra-free"],
         ["opencode-free", "deepseek-v4-flash-free"],
         ["opencode-free", "north-mini-code-free"],
+        ["opencode-free", "laguna-s-2.1-free"],
+        ["opencode-free", "ling-3.0-flash-free"],
         ["opencode-free", "big-pickle"],
-        ["opencode-free", "nemotron-3-super-free"],
-        ["opencode-free", "minimax-m2.5-free"],
         ["opencode-free", "mimo-v2.5-free"],
-        ["opencode-free", "grok-code"],
         ["github-models", "deepseek/deepseek-v3-0324"],
         ["openai", "gpt-5.6-luna"],
         ["openrouter", "deepseek/deepseek-v3.2"],
@@ -164,6 +161,15 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
     assert generated_config_match is not None
     generated_config = json.loads(generated_config_match.group(1))
     free_models = generated_config["provider"]["opencode-free"]["models"]
+    assert set(free_models) == {
+        "nemotron-3-ultra-free",
+        "deepseek-v4-flash-free",
+        "north-mini-code-free",
+        "laguna-s-2.1-free",
+        "ling-3.0-flash-free",
+        "big-pickle",
+        "mimo-v2.5-free",
+    }
     nemotron_model = free_models["nemotron-3-ultra-free"]
     deepseek_model = free_models["deepseek-v4-flash-free"]
     north_model = free_models["north-mini-code-free"]
@@ -171,9 +177,26 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
     assert nemotron_model["limit"] == {"context": 1000000, "output": 128000}
     assert "response_format" not in nemotron_model.get("options", {})
     assert deepseek_model["tool_call"] is True
+    assert deepseek_model["limit"] == {"context": 200000, "output": 128000}
     assert "response_format" not in deepseek_model.get("options", {})
     assert north_model["tool_call"] is True
     assert "response_format" not in north_model["options"]
+    assert free_models["laguna-s-2.1-free"]["limit"] == {
+        "context": 256000,
+        "output": 32000,
+    }
+    assert free_models["ling-3.0-flash-free"]["limit"] == {
+        "context": 262144,
+        "output": 32768,
+    }
+    assert free_models["big-pickle"]["limit"] == {
+        "context": 200000,
+        "output": 32000,
+    }
+    assert free_models["mimo-v2.5-free"]["limit"] == {
+        "context": 200000,
+        "output": 32000,
+    }
     assert github_candidate_models == [
         "deepseek/deepseek-v3-0324",
         "openai/gpt-4.1",
@@ -1242,7 +1265,7 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
         in workflow
     )
     assert "OpenCode model pool exceeded the outer" in workflow
-    assert 'OPENCODE_POOL_MAX_CYCLES: "0"' in workflow
+    assert 'OPENCODE_POOL_MAX_CYCLES: "1"' in workflow
     assert re.search(
         r"Run OpenCode PR Review model pool[\s\S]{0,280}continue-on-error: true",
         workflow,
@@ -1273,11 +1296,10 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
         "'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
         "opencode-free/north-mini-code-free "
+        "opencode-free/laguna-s-2.1-free "
+        "opencode-free/ling-3.0-flash-free "
         "opencode-free/big-pickle "
-        "opencode-free/nemotron-3-super-free "
-        "opencode-free/minimax-m2.5-free "
-        "opencode-free/mimo-v2.5-free "
-        "opencode-free/grok-code ' || ''"
+        "opencode-free/mimo-v2.5-free ' || ''"
     ) in workflow
     assert (
         "github-models/deepseek/deepseek-v3-0324 "
@@ -1296,7 +1318,7 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
     assert 'OPENCODE_EXPORT_TIMEOUT_SECONDS: "180"' in workflow
     assert 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "11700"' in workflow
     assert 'OPENCODE_POOL_STEP_TIMEOUT_SECONDS: "12000"' in workflow
-    assert 'OPENCODE_POOL_MAX_CYCLES: "0"' in workflow
+    assert 'OPENCODE_POOL_MAX_CYCLES: "1"' in workflow
     assert 'OPENCODE_DYNAMIC_REVIEW_CADENCE: "true"' in workflow
     assert (
         "OPENCODE_CHANGED_FILES_FILE: ${{ runner.temp }}/opencode-changed-files.txt"
@@ -1312,10 +1334,10 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
     assert 'OPENCODE_UNKNOWN_CHANGE_TOTAL_BUDGET_SECONDS: "11700"' in workflow
     assert 'OPENCODE_DYNAMIC_RUN_TIMEOUT_CAP_SECONDS: "5400"' in workflow
     assert 'OPENCODE_DYNAMIC_TOTAL_BUDGET_CAP_SECONDS: "11700"' in workflow
-    assert 'OPENCODE_DYNAMIC_MAX_CYCLES_CAP: "0"' in workflow
+    assert 'OPENCODE_DYNAMIC_MAX_CYCLES_CAP: "1"' in workflow
     assert 'OPENCODE_FREE_RUN_TIMEOUT_SECONDS: "600"' in workflow
     assert 'OPENCODE_GITHUB_GPT5_RUN_TIMEOUT_SECONDS: "45"' in workflow
-    assert 'OPENCODE_DYNAMIC_MAX_CYCLES: "0"' in workflow
+    assert 'OPENCODE_DYNAMIC_MAX_CYCLES: "1"' in workflow
     assert 'OPENCODE_BACKOFF_MAX_SECONDS: "30"' in workflow
     publish_step = workflow.split("      - name: Publish OpenCode review outcome", 1)[
         1

From 47c26e77826e4a0d8c25f5f72fa034a1edc017da Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Thu, 30 Jul 2026 05:11:36 +0000
Subject: [PATCH 4/5] ci(opencode-review): drop laguna-s-2.1-free (no
 tool_call) from Zen pool
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The bounded active free-pool set includes laguna-s-2.1-free, but the
models.dev OpenCode provider registry shows that model has no
`tool_call = true` capability. OpenCode review runs a tool-driven agent
(CodeGraph / read / evidence-gathering), so a model that cannot call tools
cannot produce a valid review verdict — it only burns a pool attempt and
falls through. Remove it from both the `opencode-free` provider block and
the public-PR `OPENCODE_MODEL_CANDIDATES` prefix, leaving six verified
free + tool_call + active reviewers (nemotron-3-ultra-free,
deepseek-v4-flash-free, north-mini-code-free, big-pickle,
ling-3.0-flash-free, mimo-v2.5-free).

Contract test updated in lockstep (candidate list, provider set, and
limit assertions); full suite green (710 passed).

Co-Authored-By: Claude <noreply@anthropic.com>
---
 .github/workflows/opencode-review-dispatch.yml | 11 +----------
 tests/test_opencode_agent_contract.py          |  9 ---------
 2 files changed, 1 insertion(+), 19 deletions(-)

diff --git a/.github/workflows/opencode-review-dispatch.yml b/.github/workflows/opencode-review-dispatch.yml
index 91ecded91..5bfd3f68a 100644
--- a/.github/workflows/opencode-review-dispatch.yml
+++ b/.github/workflows/opencode-review-dispatch.yml
@@ -3466,15 +3466,6 @@ jobs:
                       "output": 32000
                     }
                   },
-                  "laguna-s-2.1-free": {
-                    "name": "Laguna S 2.1 Free",
-                    "tool_call": true,
-                    "reasoning": true,
-                    "limit": {
-                      "context": 256000,
-                      "output": 32000
-                    }
-                  },
                   "ling-3.0-flash-free": {
                     "name": "Ling-3.0-flash Free",
                     "tool_call": true,
@@ -3824,7 +3815,7 @@ jobs:
           # cost-efficient tier, cheaper than the legacy gpt-5 it replaced
           # ($1/$6 vs $1.25/$10 per 1M tokens) so the org OpenAI budget
           # stretches further between top-ups.
-          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/laguna-s-2.1-free opencode-free/ling-3.0-flash-free opencode-free/big-pickle opencode-free/mimo-v2.5-free ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
+          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/ling-3.0-flash-free opencode-free/big-pickle opencode-free/mimo-v2.5-free ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
           # One attempt per model, then fall through to the next model. Retrying
           # the SAME model 5x let a rate-limited/hung leader consume the whole
           # step, so the pool never reached a healthy fallback model.
diff --git a/tests/test_opencode_agent_contract.py b/tests/test_opencode_agent_contract.py
index 12ffcb5ef..06e289a8a 100644
--- a/tests/test_opencode_agent_contract.py
+++ b/tests/test_opencode_agent_contract.py
@@ -95,7 +95,6 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "&& 'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
         "opencode-free/north-mini-code-free "
-        "opencode-free/laguna-s-2.1-free "
         "opencode-free/ling-3.0-flash-free "
         "opencode-free/big-pickle "
         "opencode-free/mimo-v2.5-free ' || '' }}"
@@ -106,7 +105,6 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "opencode-free/nemotron-3-ultra-free",
         "opencode-free/deepseek-v4-flash-free",
         "opencode-free/north-mini-code-free",
-        "opencode-free/laguna-s-2.1-free",
         "opencode-free/ling-3.0-flash-free",
         "opencode-free/big-pickle",
         "opencode-free/mimo-v2.5-free",
@@ -130,7 +128,6 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         ["opencode-free", "nemotron-3-ultra-free"],
         ["opencode-free", "deepseek-v4-flash-free"],
         ["opencode-free", "north-mini-code-free"],
-        ["opencode-free", "laguna-s-2.1-free"],
         ["opencode-free", "ling-3.0-flash-free"],
         ["opencode-free", "big-pickle"],
         ["opencode-free", "mimo-v2.5-free"],
@@ -165,7 +162,6 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "nemotron-3-ultra-free",
         "deepseek-v4-flash-free",
         "north-mini-code-free",
-        "laguna-s-2.1-free",
         "ling-3.0-flash-free",
         "big-pickle",
         "mimo-v2.5-free",
@@ -181,10 +177,6 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
     assert "response_format" not in deepseek_model.get("options", {})
     assert north_model["tool_call"] is True
     assert "response_format" not in north_model["options"]
-    assert free_models["laguna-s-2.1-free"]["limit"] == {
-        "context": 256000,
-        "output": 32000,
-    }
     assert free_models["ling-3.0-flash-free"]["limit"] == {
         "context": 262144,
         "output": 32768,
@@ -1296,7 +1288,6 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
         "'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
         "opencode-free/north-mini-code-free "
-        "opencode-free/laguna-s-2.1-free "
         "opencode-free/ling-3.0-flash-free "
         "opencode-free/big-pickle "
         "opencode-free/mimo-v2.5-free ' || ''"

From 03121e2492b41e9047e0f8a0caa731a80b686378 Mon Sep 17 00:00:00 2001
From: Seongho Bae <me@seonghobae.me>
Date: Thu, 30 Jul 2026 14:16:47 +0900
Subject: [PATCH 5/5] fix(review): restore catalog-verified Laguna candidate

---
 .github/workflows/opencode-review-dispatch.yml | 11 ++++++++++-
 tests/test_opencode_agent_contract.py          |  9 +++++++++
 2 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/opencode-review-dispatch.yml b/.github/workflows/opencode-review-dispatch.yml
index 5bfd3f68a..91ecded91 100644
--- a/.github/workflows/opencode-review-dispatch.yml
+++ b/.github/workflows/opencode-review-dispatch.yml
@@ -3466,6 +3466,15 @@ jobs:
                       "output": 32000
                     }
                   },
+                  "laguna-s-2.1-free": {
+                    "name": "Laguna S 2.1 Free",
+                    "tool_call": true,
+                    "reasoning": true,
+                    "limit": {
+                      "context": 256000,
+                      "output": 32000
+                    }
+                  },
                   "ling-3.0-flash-free": {
                     "name": "Ling-3.0-flash Free",
                     "tool_call": true,
@@ -3815,7 +3824,7 @@ jobs:
           # cost-efficient tier, cheaper than the legacy gpt-5 it replaced
           # ($1/$6 vs $1.25/$10 per 1M tokens) so the org OpenAI budget
           # stretches further between top-ups.
-          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/ling-3.0-flash-free opencode-free/big-pickle opencode-free/mimo-v2.5-free ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
+          OPENCODE_MODEL_CANDIDATES: "${{ needs.validate-pr-metadata.outputs.is_private == 'false' && 'opencode-free/nemotron-3-ultra-free opencode-free/deepseek-v4-flash-free opencode-free/north-mini-code-free opencode-free/laguna-s-2.1-free opencode-free/ling-3.0-flash-free opencode-free/big-pickle opencode-free/mimo-v2.5-free ' || '' }}github-models/deepseek/deepseek-v3-0324 openai/gpt-5.6-luna openrouter/deepseek/deepseek-v3.2 openrouter/qwen/qwen3-coder github-models/openai/gpt-4.1 github-models/openai/gpt-5 github-models/openai/gpt-5-chat github-models/openai/o3 github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1"
           # One attempt per model, then fall through to the next model. Retrying
           # the SAME model 5x let a rate-limited/hung leader consume the whole
           # step, so the pool never reached a healthy fallback model.
diff --git a/tests/test_opencode_agent_contract.py b/tests/test_opencode_agent_contract.py
index 06e289a8a..12ffcb5ef 100644
--- a/tests/test_opencode_agent_contract.py
+++ b/tests/test_opencode_agent_contract.py
@@ -95,6 +95,7 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "&& 'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
         "opencode-free/north-mini-code-free "
+        "opencode-free/laguna-s-2.1-free "
         "opencode-free/ling-3.0-flash-free "
         "opencode-free/big-pickle "
         "opencode-free/mimo-v2.5-free ' || '' }}"
@@ -105,6 +106,7 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "opencode-free/nemotron-3-ultra-free",
         "opencode-free/deepseek-v4-flash-free",
         "opencode-free/north-mini-code-free",
+        "opencode-free/laguna-s-2.1-free",
         "opencode-free/ling-3.0-flash-free",
         "opencode-free/big-pickle",
         "opencode-free/mimo-v2.5-free",
@@ -128,6 +130,7 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         ["opencode-free", "nemotron-3-ultra-free"],
         ["opencode-free", "deepseek-v4-flash-free"],
         ["opencode-free", "north-mini-code-free"],
+        ["opencode-free", "laguna-s-2.1-free"],
         ["opencode-free", "ling-3.0-flash-free"],
         ["opencode-free", "big-pickle"],
         ["opencode-free", "mimo-v2.5-free"],
@@ -162,6 +165,7 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
         "nemotron-3-ultra-free",
         "deepseek-v4-flash-free",
         "north-mini-code-free",
+        "laguna-s-2.1-free",
         "ling-3.0-flash-free",
         "big-pickle",
         "mimo-v2.5-free",
@@ -177,6 +181,10 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
     assert "response_format" not in deepseek_model.get("options", {})
     assert north_model["tool_call"] is True
     assert "response_format" not in north_model["options"]
+    assert free_models["laguna-s-2.1-free"]["limit"] == {
+        "context": 256000,
+        "output": 32000,
+    }
     assert free_models["ling-3.0-flash-free"]["limit"] == {
         "context": 262144,
         "output": 32768,
@@ -1288,6 +1296,7 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
         "'opencode-free/nemotron-3-ultra-free "
         "opencode-free/deepseek-v4-flash-free "
         "opencode-free/north-mini-code-free "
+        "opencode-free/laguna-s-2.1-free "
         "opencode-free/ling-3.0-flash-free "
         "opencode-free/big-pickle "
         "opencode-free/mimo-v2.5-free ' || ''"