Remove workflow-only deterministic approval fallback

.github/workflows/opencode-review.yml

@@ -4325,32 +4325,6 @@ jobs:
               changed_files_summary="bounded current-head evidence"
             fi
 
-            if low_risk_model_exhaustion_fallback_applies; then
-              body="$(printf '%s\n' \
-                "## Pull request overview" \
-                "" \
-                "OpenCode exhausted the configured model pool without a usable current-head review conclusion, but this PR only changes workflow files and all current-head peer gates passed." \
-                "" \
-                "## Findings" \
-                "" \
-                "No blocking findings." \
-                "" \
-                "## Summary" \
-                "" \
-                "- Result: APPROVE" \
-                "- Reason: model output was unavailable, but deterministic current-head gates passed for a workflow-only change." \
-                "- Deterministic evidence: current-head changed-file evidence (${changed_files_summary}); coverage-evidence result ${COVERAGE_EVIDENCE_RESULT:-unknown}; peer checks from statusCheckRollup excluding this OpenCode check; no unresolved human review threads; mergeability was not dirty." \
-                "- Model outcome: model_pool=${OPENCODE_MODEL_POOL_OUTCOME:-unknown}; selected_model=${OPENCODE_MODEL_POOL_MODEL:-none}." \
-                "- Head SHA: \`${HEAD_SHA}\`" \
-                "- Workflow run: ${RUN_ID}" \
-                "- Workflow attempt: ${RUN_ATTEMPT}" \
-                "" \
-                "This bounded workflow-only fallback does not apply to source, test, lockfile, dependency manifest, generated artifact, or documentation changes.")"
-              body="$(ensure_review_body_has_change_graph "$body")"
-              create_pull_review "APPROVE" "$body"
-              return 0
-            fi
-
             first_line_file="$(mktemp)"
             if gh pr diff "$PR_NUMBER" --repo "$GH_REPOSITORY" --patch 2>/dev/null |
               awk '
@@ -4434,26 +4408,6 @@ jobs:
             return 0
           }
 
-          low_risk_model_exhaustion_fallback_applies() {
-            local changed_count=0
-            local changed_path
-
-            if [ ! -s "${OPENCODE_CHANGED_FILES_FILE:-}" ]; then
-              return 1
-            fi
-
-            while IFS= read -r changed_path; do
-              [ -n "$changed_path" ] || continue
-              changed_count=$((changed_count + 1))
-              case "$changed_path" in
-                .github/workflows/*.yml|.github/workflows/*.yaml) ;;
-                *) return 1 ;;
-              esac
-            done <"$OPENCODE_CHANGED_FILES_FILE"
-
-            [ "$changed_count" -gt 0 ] && [ "$changed_count" -le 2 ]
-          }
-
                     request_changes_for_merge_conflict_if_present() {
                       local pr_json merge_state mergeable base_ref head_ref body change_graph
 

scripts/ci/run_opencode_review_model_pool.sh

@@ -52,6 +52,39 @@ write_prompt() {
 		python3 "$GITHUB_WORKSPACE/scripts/ci/render_opencode_prompt_template.py" "$prompt_file"
 }
 
+reasoning_capable_model() {
+	case "$1" in
+	openai/gpt-5* | openai/o3* | openai/o4* | deepseek/deepseek-r1*)
+		return 0
+		;;
+	*)
+		return 1
+		;;
+	esac
+}
+
+assert_reasoning_effort_for_candidate() {
+	local model_candidate="$1"
+	local provider="${model_candidate%%/*}"
+	local model_name="${model_candidate#*/}"
+
+	if ! reasoning_capable_model "$model_name"; then
+		return 0
+	fi
+	if [ "$provider" = "$model_candidate" ] || [ -z "$model_name" ]; then
+		printf 'OpenCode candidate %s is not provider-qualified.\n' "$model_candidate"
+		return 1
+	fi
+	if ! jq -e --arg provider "$provider" --arg model "$model_name" '
+		.provider[$provider].models[$model].reasoning == true
+		and .provider[$provider].models[$model].options.reasoningEffort == "high"
+		and .provider[$provider].models[$model].variants.high.reasoningEffort == "high"
+	' opencode.jsonc >/dev/null; then
+		printf 'OpenCode reasoning-capable candidate %s must set reasoningEffort=high in opencode.jsonc.\n' "$model_candidate"
+		return 1
+	fi
+}
+
 run_one_model_attempt() {
 	local model_candidate="$1"
 	local attempt="$2"
@@ -115,6 +148,7 @@ main() {
 	cd "$OPENCODE_REVIEW_WORKDIR"
 
 	for model_candidate in $OPENCODE_MODEL_CANDIDATES; do
+		assert_reasoning_effort_for_candidate "$model_candidate"
 		safe_model="${model_candidate//\//-}"
 		prompt_file="${RUNNER_TEMP}/opencode-review-${safe_model}-prompt.md"
 		candidate_output_file="${RUNNER_TEMP}/opencode-review-${safe_model}.md"

scripts/ci/test_strix_quick_gate.sh

@@ -502,6 +502,8 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
 	assert_file_contains "$workflow_file" "Do not spend the session listing every changed path before reviewing" "opencode review prompt prevents fallback sessions from exhausting steps on file listing"
 	assert_file_contains "$workflow_file" "Always return a final control block instead of a progress summary" "opencode review prompt requires a gate conclusion instead of a progress summary"
 	assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" 'timeout --kill-after=30s "${run_timeout_seconds}s" opencode run' "opencode review model pool has a kill-after bounded timeout"
+	assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" "assert_reasoning_effort_for_candidate" "opencode review validates high reasoning effort before running capable model candidates"
+	assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" 'reasoningEffort == "high"' "opencode review requires high reasoning effort in opencode.jsonc for capable models"
 	assert_file_contains "$workflow_file" 'OPENCODE_RUN_TIMEOUT_SECONDS: "180"' "opencode primary review is bounded tightly enough to reach fallback models promptly"
 	assert_file_contains "$workflow_file" 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "2400"' "opencode model pool has a merge-safe total retry budget"
 	assert_file_contains "$workflow_file" "needs.coverage-evidence.result == 'success'" "opencode model pool only runs after coverage evidence passed"
@@ -578,8 +580,8 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
 	assert_file_not_contains "$workflow_file" 'approve_review_tooling_bootstrap_after_model_failure' "opencode approval must not use deterministic review-tooling bootstrap approval after model-output failures"
 	assert_file_not_contains "$workflow_file" 'Deterministic review-tooling bootstrap fallback approval was used' "opencode approval must not publish model-exhaustion approvals"
 	assert_file_not_contains "$workflow_file" 'model-exhaustion approval did not apply' "opencode approval failure text should describe retry exhaustion, not model-exhaustion criteria"
-	assert_file_contains "$workflow_file" "low_risk_model_exhaustion_fallback_applies" "opencode approval has a narrow workflow-only fallback after green current-head gates"
-	assert_file_contains "$workflow_file" "This bounded workflow-only fallback does not apply to source, test, lockfile, dependency manifest, generated artifact, or documentation changes." "opencode workflow-only fallback documents its excluded surfaces"
+	assert_file_not_contains "$workflow_file" "low_risk_model_exhaustion_fallback_applies" "opencode approval must not approve from workflow-only deterministic fallback"
+	assert_file_not_contains "$workflow_file" "This bounded workflow-only fallback does not apply to source, test, lockfile, dependency manifest, generated artifact, or documentation changes." "opencode approval must not publish model-exhaustion approvals"
 	assert_file_contains "$workflow_file" 'wait_for_peer_github_checks "$pending_file"' "model-failure hold waits for peer checks before changing review state"
 	assert_file_contains "$workflow_file" 'pending_checks_file="$(mktemp)"' "model-failure hold writes pending-check evidence to a real temp file"
 	assert_file_contains "$workflow_file" 'collect_github_checks_with_retry collect_failed_github_checks "$failed_file"' "model-failure hold rejects current-head failed peer checks"
@@ -602,8 +604,8 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
 	assert_file_contains "$workflow_file" 'scripts/ci/test_strix_quick_gate.sh)' "opencode central review fallback allowlist includes only the central gate self-test"
 	assert_file_contains "$workflow_file" 'wait_for_peer_github_checks "$pending_checks_file"' "opencode model-failure path waits for peer checks before failing closed"
 	assert_file_contains "$workflow_file" 'collect_unresolved_human_review_threads "$unresolved_human_threads_file"' "opencode model-failure path re-queries human threads before failing closed"
-	assert_file_contains "$workflow_file" ".github/workflows/*.yml|.github/workflows/*.yaml" "opencode model-exhaustion fallback is limited to workflow files"
-	assert_file_contains "$workflow_file" '[ "$changed_count" -gt 0 ] && [ "$changed_count" -le 2 ]' "opencode workflow-only fallback caps changed-file count"
+	assert_file_not_contains "$workflow_file" ".github/workflows/*.yml|.github/workflows/*.yaml" "opencode model-exhaustion fallback must not allow workflow-only deterministic approval"
+	assert_file_not_contains "$workflow_file" '[ "$changed_count" -gt 0 ] && [ "$changed_count" -le 2 ]' "opencode model-exhaustion fallback must not cap deterministic approval scope"
 	assert_file_contains "$workflow_file" "all configured OpenCode model attempts failed to produce a usable current-head control block" "opencode model-output failures fail the check without publishing a review"
 	assert_file_contains "$workflow_file" "no model produced a valid review control block" "opencode model-failure path documents why approval is withheld"
 	assert_file_contains "$workflow_file" 'OPENCODE_MODEL_ATTEMPTS: "2"' "opencode primary and deepseek review paths retry model execution"
@@ -857,7 +859,7 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
 	assert_file_contains "$REPO_ROOT/scripts/ci/emit_opencode_failed_check_fallback_findings.sh" '^CMD \["/app/scripts/docker_entrypoint\.sh"\]' "opencode failed-check fallback maps missing Docker entrypoint reports to the Dockerfile CMD line"
 	assert_file_contains "$workflow_file" "Unrelated speculative findings are invalid when failed-check evidence is present." "opencode review prompt forbids unrelated failed-check findings"
 	assert_file_contains "$workflow_file" "run_failed_check_diagnosis" "opencode approval gate reruns OpenCode diagnosis when checks fail after the initial review"
-	assert_file_contains "$workflow_file" "deterministic current-head gates passed for a workflow-only change" "opencode approval gate records the narrow workflow-only model-failure fallback"
+	assert_file_not_contains "$workflow_file" "deterministic current-head gates passed for a workflow-only change" "opencode approval gate must not record deterministic model-failure approval"
 	assert_file_contains "$workflow_file" "request_changes_after_model_exhaustion" "opencode model-failure path reuses green current-head gates only to decide fail-closed diagnostics"
 	assert_file_contains "$workflow_file" "request_changes_for_merge_conflict_if_present" "opencode approval gate checks mergeability before approving model or fallback output"
 	assert_file_contains "$workflow_file" "Merge Conflict Guidance" "opencode approval gate emits explicit conflict guidance when mergeability is dirty"

tests/test_opencode_agent_contract.py

@@ -156,6 +156,10 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
     assert "opencode_review_model_pool" in workflow
     assert "run_opencode_review_model_pool.sh" in workflow
     assert "OPENCODE_MODEL_CANDIDATES" in workflow
+    model_pool_runner = Path("scripts/ci/run_opencode_review_model_pool.sh").read_text(encoding="utf-8")
+    assert "assert_reasoning_effort_for_candidate" in model_pool_runner
+    assert 'reasoningEffort == "high"' in model_pool_runner
+    assert "OpenCode reasoning-capable candidate %s must set reasoningEffort=high" in model_pool_runner
     assert 'timeout-minutes: 45' in workflow
     assert 'OPENCODE_RUN_TIMEOUT_SECONDS: "180"' in workflow
     assert 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "2400"' in workflow