@mmerterden/multi-agent-pipeline 12.6.0 → 12.8.0

package/pipeline/scripts/smoke-diff-risk.sh

@@ -1,190 +0,0 @@
-#!/usr/bin/env bash
-# smoke-diff-risk.sh  -  v8.3.0
-#
-# Verifies the Phase 4 advisory diff-risk pipeline:
-#   1. diff-risk-score.mjs runs against the iOS fixture and produces valid JSON
-#   2. KeychainStore (security path) ranks first
-#   3. Tests file ranks last (no security/public_api/migration signal)
-#   4. validate-diff-risk.mjs accepts the output, rejects malformed
-#   5. Android fixture: migration outranks AuthRepository outranks HomeScreen
-#   6. --top truncation honored
-#   7. Empty diff returns exit 2
-#   8. phase-4-review.md ref doc declares Step 1.75 + diff-risk-score.mjs
-#   9. code-reviewer.md agent template carries the priority-files placeholder
-#   10. prefs.schema.json exposes diffRisk advisory toggle
-#   11. test-removal fixture fires the test_lines_removed signal (v1.1.0)
-#
-# Exit 0 = all pass, 1 = any failure.
-
-set -euo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-SCORE="$ROOT/pipeline/scripts/diff-risk-score.mjs"
-VALIDATE="$ROOT/pipeline/scripts/validate-diff-risk.mjs"
-SCHEMA="$ROOT/pipeline/schemas/diff-risk.schema.json"
-PHASE4="$ROOT/pipeline/multi-agent-refs/phases/phase-4-review.md"
-REVIEWER="$ROOT/pipeline/agents/code-reviewer.md"
-PREFS="$ROOT/pipeline/schemas/prefs.schema.json"
-FIX_IOS="$ROOT/pipeline/scripts/fixtures/diff-risk-ios.diff"
-FIX_AND="$ROOT/pipeline/scripts/fixtures/diff-risk-android.diff"
-FIX_TESTRM="$ROOT/pipeline/scripts/fixtures/diff-risk-test-removal.diff"
-
-pass=0
-fail=0
-failures=()
-record_pass() { pass=$((pass + 1)); printf '  \033[0;32mPASS\033[0m %s\n' "$1"; }
-record_fail() { fail=$((fail + 1)); failures+=("$1"); printf '  \033[0;31mFAIL\033[0m %s\n' "$1"; }
-
-printf '→ smoke-diff-risk (v8.3.0): pre-review risk scoring contract\n'
-
-[ -f "$SCHEMA" ]   || { record_fail "schema missing: $SCHEMA"; exit 1; }
-[ -f "$FIX_IOS" ]  || { record_fail "fixture missing: $FIX_IOS"; exit 1; }
-[ -f "$FIX_AND" ]  || { record_fail "fixture missing: $FIX_AND"; exit 1; }
-[ -f "$FIX_TESTRM" ] || { record_fail "fixture missing: $FIX_TESTRM"; exit 1; }
-
-# --- 1: iOS fixture produces JSON ---
-out_ios=$(node "$SCORE" --diff "$FIX_IOS" 2>/dev/null)
-if jq -e '.schemaVersion == "1.1.0"' <<< "$out_ios" >/dev/null 2>&1; then
-  record_pass "iOS fixture renders schema-versioned JSON"
-else
-  record_fail "iOS fixture JSON malformed or missing schemaVersion"
-fi
-
-# --- 2: KeychainStore ranks first ---
-top1=$(jq -r '.files[0].path' <<< "$out_ios")
-if [ "$top1" = "MyApp/Sources/Auth/KeychainStore.swift" ]; then
-  record_pass "iOS top-1 = KeychainStore (security_path + public_api + no_test)"
-else
-  record_fail "iOS top-1 should be KeychainStore, got: $top1"
-fi
-
-# --- 3: Test file ranks last ---
-last=$(jq -r '.files[-1].path' <<< "$out_ios")
-if [ "$last" = "MyAppTests/SettingsViewTests.swift" ]; then
-  record_pass "iOS last = test file (no risk signals beyond loc_changed)"
-else
-  record_fail "iOS last should be test file, got: $last"
-fi
-
-# --- 4: Validator accepts valid output, rejects malformed ---
-set +e
-echo "$out_ios" | node "$VALIDATE" - >/dev/null 2>&1
-rc_ok=$?
-echo '{"schemaVersion":"0.0.0","files":[]}' | node "$VALIDATE" - >/dev/null 2>&1
-rc_bad=$?
-set -e
-if [ "$rc_ok" -eq 0 ]; then
-  record_pass "validator accepts valid output"
-else
-  record_fail "validator rejected valid output (rc=$rc_ok)"
-fi
-if [ "$rc_bad" -ne 0 ]; then
-  record_pass "validator rejects malformed input"
-else
-  record_fail "validator should reject malformed (got rc=$rc_bad)"
-fi
-
-# --- 5: Android ranking ---
-out_and=$(node "$SCORE" --diff "$FIX_AND" 2>/dev/null)
-top1=$(jq -r '.files[0].path' <<< "$out_and")
-top2=$(jq -r '.files[1].path' <<< "$out_and")
-top3=$(jq -r '.files[2].path' <<< "$out_and")
-if [[ "$top1" == *"Migration_3_to_4.kt"* ]]; then
-  record_pass "Android top-1 = migration file"
-else
-  record_fail "Android top-1 should be migration, got: $top1"
-fi
-if [[ "$top2" == *"AuthRepository.kt"* ]]; then
-  record_pass "Android top-2 = AuthRepository (security_path)"
-else
-  record_fail "Android top-2 should be AuthRepository, got: $top2"
-fi
-if [[ "$top3" == *"HomeScreen.kt"* ]]; then
-  record_pass "Android top-3 = HomeScreen (ui_critical)"
-else
-  record_fail "Android top-3 should be HomeScreen, got: $top3"
-fi
-
-# --- 6: --top N truncates ---
-out_top2=$(node "$SCORE" --diff "$FIX_IOS" --top 2 2>/dev/null)
-n=$(jq '.files | length' <<< "$out_top2")
-if [ "$n" = "2" ]; then
-  record_pass "--top 2 truncates files array"
-else
-  record_fail "--top 2 should keep 2 rows, got $n"
-fi
-
-# --- 7: Empty diff exits 2 ---
-empty=$(mktemp)
-set +e
-node "$SCORE" --diff "$empty" >/dev/null 2>&1
-rc=$?
-set -e
-rm -f "$empty"
-if [ "$rc" -eq 2 ]; then
-  record_pass "empty diff exits 2"
-else
-  record_fail "empty diff should exit 2, got $rc"
-fi
-
-# --- 8: phase-4-review.md declares the advisory step ---
-if grep -qE '^####? .*Diff Risk' "$PHASE4"; then
-  record_pass "phase-4-review.md declares Diff Risk step"
-else
-  record_fail "phase-4-review.md missing Diff Risk step heading"
-fi
-if grep -q 'diff-risk-score.mjs' "$PHASE4"; then
-  record_pass "phase-4-review.md references diff-risk-score.mjs"
-else
-  record_fail "phase-4-review.md missing script reference"
-fi
-
-# --- 9: code-reviewer.md template carries priority placeholder ---
-if grep -q 'Priority Files' "$REVIEWER"; then
-  record_pass "code-reviewer.md declares Priority Files section"
-else
-  record_fail "code-reviewer.md missing Priority Files section"
-fi
-
-# --- 10: prefs schema exposes diffRisk advisory toggle ---
-if jq -e '.properties.global.properties.diffRiskAdvisory' "$PREFS" >/dev/null 2>&1; then
-  record_pass "prefs schema exposes diffRiskAdvisory toggle"
-else
-  record_fail "prefs.schema.json missing global.diffRiskAdvisory"
-fi
-
-# --- 11: test_lines_removed signal fires on the test-removal fixture ---
-out_testrm=$(node "$SCORE" --diff "$FIX_TESTRM" 2>/dev/null)
-sig_value=$(jq -r '.files[] | select(.path == "MyAppTests/LoginViewModelTests.swift")
-                   | .signals[] | select(.name == "test_lines_removed") | .value' <<< "$out_testrm")
-if [ "$sig_value" = "16" ]; then
-  record_pass "test_lines_removed fires with value=16 (18 removed - 2 added)"
-else
-  record_fail "test_lines_removed should fire with value=16, got: ${sig_value:-missing}"
-fi
-sig_on_source=$(jq -r '[.files[] | select(.path == "MyApp/Sources/Auth/LoginViewModel.swift")
-                        | .signals[] | select(.name == "test_lines_removed")] | length' <<< "$out_testrm")
-if [ "$sig_on_source" = "0" ]; then
-  record_pass "test_lines_removed does not fire on source files"
-else
-  record_fail "test_lines_removed must only fire on test-classified paths"
-fi
-set +e
-echo "$out_testrm" | node "$VALIDATE" - >/dev/null 2>&1
-rc_testrm=$?
-set -e
-if [ "$rc_testrm" -eq 0 ]; then
-  record_pass "validator accepts output carrying test_lines_removed"
-else
-  record_fail "validator rejected test_lines_removed output (rc=$rc_testrm)"
-fi
-
-# --- Summary ---
-total=$((pass + fail))
-printf '\n→ smoke-diff-risk: %d/%d passed\n' "$pass" "$total"
-if [ "$fail" -ne 0 ]; then
-  printf '\nFailures:\n'
-  for f in "${failures[@]}"; do printf '  - %s\n' "$f"; done
-  exit 1
-fi
-exit 0

package/pipeline/scripts/smoke-dynamic-skill-loading.sh

@@ -1,160 +0,0 @@
-#!/usr/bin/env bash
-# smoke-dynamic-skill-loading.sh  -  v7.0.I
-#
-# Verifies the index + matcher contract:
-#   1. build-skills-index.mjs produces valid JSON with schemaVersion + entries
-#   2. index skill count matches the raw filesystem count
-#   3. match-skills.mjs ranks iOS skills higher on a SwiftUI task
-#   4. match-skills.mjs --json returns sorted matches[]
-#   5. matcher platform boost applies when --stack matches frontmatter
-#   6. matcher returns empty matches[] for a task with no relevant skills
-#   7. matcher --limit honored
-#   8. prefs schema exposes dynamicSkillLoading default false
-#
-# Exit 0 all pass, 1 any failure.
-
-set -euo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-BUILD="$ROOT/pipeline/scripts/build-skills-index.mjs"
-MATCH="$ROOT/pipeline/scripts/match-skills.mjs"
-SCHEMA="$ROOT/pipeline/schemas/prefs.schema.json"
-
-pass=0; fail=0; failures=()
-record_pass() { pass=$((pass + 1)); printf '  \033[0;32mPASS\033[0m %s\n' "$1"; }
-record_fail() { fail=$((fail + 1)); failures+=("$1"); printf '  \033[0;31mFAIL\033[0m %s\n' "$1"; }
-
-printf '→ smoke-dynamic-skill-loading (v7.0.I): index + matcher contract\n'
-
-# Build into a sandbox so we don't clobber the shipping index
-sandbox=$(mktemp -d)
-skills="$sandbox/skills"
-mkdir -p "$skills/core/multi-agent" "$skills/core/testing-backend" "$skills/external/swiftui-pro" "$skills/external/compose-components"
-
-cat > "$skills/core/multi-agent/SKILL.md" <<'S'
----
-name: multi-agent
-description: Task orchestrator for the multi-agent pipeline
-platform: generic
----
-body
-S
-
-cat > "$skills/core/testing-backend/SKILL.md" <<'S'
----
-name: testing-backend
-description: Backend testing patterns for pytest / Jest
-platform: backend
-trigger-keywords: pytest, jest, integration test
----
-body
-S
-
-cat > "$skills/external/swiftui-pro/SKILL.md" <<'S'
----
-name: swiftui-pro
-description: SwiftUI review for modern APIs, state management, performance
-platform: ios
-trigger-keywords: swiftui, swift, ios, swiftui performance
-trigger-paths: *.swift, **/Sources/**/*.swift
----
-body
-S
-
-cat > "$skills/external/compose-components/SKILL.md" <<'S'
----
-name: compose-components
-description: Material 3 Jetpack Compose components
-platform: android
-trigger-keywords: compose, jetpack, material 3, android
-trigger-paths: *.kt, **/src/main/**/*.kt
----
-body
-S
-
-# --- 1: build index ---
-node "$BUILD" --root "$skills" >/dev/null
-INDEX="$skills/.skills-index.json"
-if [ -f "$INDEX" ] && jq -e '.schemaVersion and .entries and (.skillCount == 4)' "$INDEX" >/dev/null; then
-  record_pass "build-skills-index emits valid JSON with 4 entries"
-else
-  record_fail "build-skills-index output malformed"
-fi
-
-# --- 2: count matches filesystem ---
-raw_count=$(find "$skills" -name SKILL.md | wc -l | tr -d ' ')
-index_count=$(jq -r '.skillCount' "$INDEX")
-if [ "$raw_count" = "$index_count" ]; then
-  record_pass "index count matches filesystem ($index_count)"
-else
-  record_fail "index count $index_count != filesystem $raw_count"
-fi
-
-# --- 3: iOS task ranks SwiftUI skills higher ---
-out=$(node "$MATCH" "SwiftUI dark mode fix on LoginView" \
-  --touched-files "Sources/Auth/LoginView.swift" \
-  --stack ios --limit 4 --index "$INDEX" --json)
-top=$(jq -r '.matches[0].name' <<< "$out")
-if [ "$top" = "swiftui-pro" ]; then
-  record_pass "iOS task ranks swiftui-pro top ($top)"
-else
-  record_fail "iOS task should rank swiftui-pro first, got $top"
-fi
-
-# --- 4: JSON output is sorted ---
-scores=$(jq -r '.matches[].score' <<< "$out")
-prev=9999; sorted=1
-for s in $scores; do
-  if [ "$s" -gt "$prev" ]; then sorted=0; break; fi
-  prev=$s
-done
-if [ "$sorted" = "1" ]; then
-  record_pass "JSON matches[] sorted by score desc"
-else
-  record_fail "matches[] not sorted"
-fi
-
-# --- 5: platform boost ---
-# A keyword-free task with only --stack android should still boost Android skill
-out=$(node "$MATCH" "Build the feature end to end" --stack android --limit 4 --index "$INDEX" --json)
-rules=$(jq -r '.matches[] | select(.name=="compose-components") | .reasons[].rule' <<< "$out" 2>/dev/null || echo "")
-if echo "$rules" | grep -q 'platform'; then
-  record_pass "platform boost reason present for matched stack"
-else
-  record_fail "platform boost missing for --stack android"
-fi
-
-# --- 6: empty matches ---
-out=$(node "$MATCH" "Totally unrelated lmn xyz qrs foobar baz" --limit 4 --index "$INDEX" --json)
-n=$(jq '.matches | length' <<< "$out")
-if [ "$n" = "0" ]; then
-  record_pass "empty matches[] for unrelated task"
-else
-  record_fail "expected 0 matches for unrelated task, got $n"
-fi
-
-# --- 7: --limit honored ---
-out=$(node "$MATCH" "swiftui compose test pipeline" --limit 2 --index "$INDEX" --json)
-n=$(jq '.matches | length' <<< "$out")
-if [ "$n" -le "2" ]; then
-  record_pass "--limit 2 returns at most 2 matches ($n)"
-else
-  record_fail "--limit 2 returned $n"
-fi
-
-# --- 8: schema pref ---
-if jq -e '.properties.global.properties.dynamicSkillLoading.default == false' "$SCHEMA" >/dev/null 2>&1; then
-  record_pass "prefs schema exposes dynamicSkillLoading default false"
-else
-  record_fail "prefs schema missing dynamicSkillLoading or wrong default"
-fi
-
-rm -rf "$sandbox"
-
-printf '\n══ dynamic-skill-loading smoke: %d passed, %d failed ══\n' "$pass" "$fail"
-if [ "$fail" -gt 0 ]; then
-  printf '\nFailures:\n'
-  for m in "${failures[@]}"; do printf '  - %s\n' "$m"; done
-  exit 1
-fi
-exit 0

package/pipeline/scripts/smoke-eval-live.sh

@@ -1,136 +0,0 @@
-#!/usr/bin/env bash
-# smoke-eval-live.sh  -  v7.8.0 Paket A: opt-in live golden-task runner.
-#
-# Validates the cost-guarding contract WITHOUT making any model calls. Live
-# mode is gated behind two independent signals (--live flag AND
-# MULTI_AGENT_LIVE_EVAL=1 env) so accidental cost is impossible during CI
-# or local `npm test` runs.
-#
-# All assertions exercise the dry-run code path or expect early-exit before
-# any model invocation.
-
-set -uo pipefail
-
-REPO_ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-SCRIPT="$REPO_ROOT/pipeline/scripts/eval-golden-tasks-live.mjs"
-
-PASS=0
-FAIL=0
-pass() { PASS=$((PASS + 1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL + 1)); echo "  ✗ $1"; }
-
-cleanup() { [ -n "${TMP:-}" ] && rm -rf "$TMP"; }
-trap cleanup EXIT
-TMP=$(mktemp -d)
-
-# Strip env that would unlock live mode; smoke MUST never spend money.
-unset MULTI_AGENT_LIVE_EVAL
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 1. Script presence + syntax"
-[ -x "$SCRIPT" ] && pass "eval-golden-tasks-live.mjs is executable" || fail "script not executable"
-node --check "$SCRIPT" 2>/dev/null && pass "syntax check" || fail "syntax error"
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 2. Default mode is dry-run (no --live, no env)"
-OUT=$(node "$SCRIPT" --json 2>/dev/null)
-if echo "$OUT" | grep -q '"mode": "dry-run"'; then
-  pass "default mode = dry-run"
-else
-  fail "default mode not dry-run: $(echo "$OUT" | head -3)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 3. Dry-run does not invoke model"
-# All cases must end with status "skipped-dry-run"  -  i.e. no live call
-SKIPPED_COUNT=$(echo "$OUT" | grep -c '"pass": "skipped-dry-run"' || true)
-if [ "$SKIPPED_COUNT" -ge 1 ]; then
-  pass "dry-run skipped all cases ($SKIPPED_COUNT) without invoking model"
-else
-  fail "dry-run did not skip cases properly"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 4. --live without env exits 2 (cost gate)"
-node "$SCRIPT" --live 2>"$TMP/err.txt"
-EC=$?
-if [ "$EC" -eq 2 ] && grep -q "MULTI_AGENT_LIVE_EVAL" "$TMP/err.txt"; then
-  pass "--live without env: exit $EC + cost gate message"
-else
-  fail "--live cost gate weak (exit=$EC, err='$(cat "$TMP/err.txt")')"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 5. Static check  -  no fall-through past cost gate"
-# The cost gate must be the first thing checked after flag parsing. If --live
-# is set without env, no model call should be reachable. Verify by static
-# grep: spawnSync("claude", ...) must come AFTER the env-check block.
-GATE_LINE=$(grep -n 'MULTI_AGENT_LIVE_EVAL' "$SCRIPT" | head -1 | cut -d: -f1)
-CALL_LINE=$(grep -n 'spawnSync("claude"' "$SCRIPT" | head -1 | cut -d: -f1)
-if [ -n "$GATE_LINE" ] && [ -n "$CALL_LINE" ] && [ "$GATE_LINE" -lt "$CALL_LINE" ]; then
-  pass "cost gate at line $GATE_LINE precedes claude call at line $CALL_LINE"
-else
-  fail "cost gate ordering wrong (gate=$GATE_LINE, call=$CALL_LINE)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 6. Budget cap honored  -  total-budget < single case budget skips all"
-OUT=$(node "$SCRIPT" --max-cases=5 --budget=0.5 --total-budget=0.1 --json 2>/dev/null)
-RUN_COUNT=$(echo "$OUT" | grep -oE '"cases_run":[[:space:]]*[0-9]+' | grep -oE '[0-9]+')
-SKIPPED=$(echo "$OUT" | grep -c '"skipped-budget-exhausted"' || true)
-if [ "$SKIPPED" -ge 1 ]; then
-  pass "budget exhaustion produces 'skipped-budget-exhausted' result ($SKIPPED times)"
-else
-  pass "budget very tight: cases_run=$RUN_COUNT (acceptable in dry-run mode)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 7. --max-cases caps the run"
-OUT=$(node "$SCRIPT" --max-cases=1 --json 2>/dev/null)
-RUN_COUNT=$(echo "$OUT" | grep -oE '"cases_run":[[:space:]]*[0-9]+' | grep -oE '[0-9]+')
-if [ "$RUN_COUNT" = "1" ]; then
-  pass "--max-cases=1 limits cases_run to 1"
-else
-  fail "--max-cases not honored: cases_run=$RUN_COUNT"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 8. --case selector"
-OUT=$(node "$SCRIPT" --case 01-ios-bugfix-darkmode --json 2>/dev/null)
-if echo "$OUT" | grep -q '"case": "01-ios-bugfix-darkmode"'; then
-  pass "--case selector picks the right fixture"
-else
-  fail "--case selector did not target fixture"
-fi
-
-node "$SCRIPT" --case nonexistent-case 2>"$TMP/err.txt"
-EC=$?
-if [ "$EC" -eq 2 ] && grep -q "not found" "$TMP/err.txt"; then
-  pass "unknown --case exits 2 with 'not found'"
-else
-  fail "unknown --case error weak (exit=$EC)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 9. Help text mentions cost gate prominently"
-HELP=$(node "$SCRIPT" --help 2>&1)
-if echo "$HELP" | grep -q "MULTI_AGENT_LIVE_EVAL"; then
-  pass "--help documents the cost gate env var"
-else
-  fail "--help missing cost gate documentation"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 10. Bad --budget exits 2"
-node "$SCRIPT" --budget=invalid 2>/dev/null
-EC=$?
-if [ "$EC" -eq 2 ]; then
-  pass "invalid --budget exits 2"
-else
-  fail "invalid --budget did not error (exit=$EC)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo ""
-echo "══ eval-live smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]

package/pipeline/scripts/smoke-evidence-gate.sh

@@ -1,93 +0,0 @@
-#!/usr/bin/env bash
-# smoke-evidence-gate.sh  -  contract check for the default-FAIL evidence gate.
-#
-# The gate must fail CLOSED: a "passed" claim with no / empty / contradicting
-# evidence is blocked; only a log that actually shows success lets the claim
-# through. Non-pass statuses pass through untouched.
-#
-# Exit 0 = all pass, 1 = any failure.
-
-set -uo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-GATE="$ROOT/pipeline/scripts/evidence-gate.mjs"
-TMP="$(mktemp -d)"
-trap 'rm -rf "$TMP"' EXIT
-
-PASS=0
-FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-
-command -v node >/dev/null 2>&1 || { echo "error: node required" >&2; exit 127; }
-[ -f "$GATE" ] || { echo "FAIL: evidence-gate.mjs missing" >&2; exit 1; }
-
-printf '** BUILD SUCCEEDED **\nCompiling 12 files\n' > "$TMP/build-ok.log"
-printf '** BUILD FAILED **\nerror: cannot find type Foo\n' > "$TMP/build-fail.log"
-printf 'Test Suite All passed\n0 failures\n' > "$TMP/test-ok.log"
-printf 'Test Suite failed\n3 failures\n' > "$TMP/test-fail.log"
-: > "$TMP/empty.log"
-
-expect_exit() {
-  local want="$1"; shift
-  local desc="$1"; shift
-  local got=0
-  node "$GATE" "$@" >/dev/null 2>&1 || got=$?
-  if [ "$got" = "$want" ]; then
-    pass "$desc (exit=$got)"
-  else
-    fail "$desc (expected $want, got $got)"
-  fi
-}
-
-echo "→ 1. Substantiated pass claims (exit 0)"
-expect_exit 0 "build passed + SUCCEEDED log" --claim build --status passed --evidence "$TMP/build-ok.log"
-expect_exit 0 "test passed + passing log" --claim test --status passed --evidence "$TMP/test-ok.log"
-
-echo ""
-echo "→ 2. Default-FAIL: unverified pass claims (exit 1)"
-expect_exit 1 "build passed + no evidence flag" --claim build --status passed
-expect_exit 1 "build passed + missing file" --claim build --status passed --evidence "$TMP/nope.log"
-expect_exit 1 "build passed + empty log" --claim build --status passed --evidence "$TMP/empty.log"
-expect_exit 1 "build passed + FAILED log (contradicted)" --claim build --status passed --evidence "$TMP/build-fail.log"
-expect_exit 1 "test passed + failing log (contradicted)" --claim test --status passed --evidence "$TMP/test-fail.log"
-
-echo ""
-echo "→ 3. Non-pass statuses pass through (exit 0)"
-expect_exit 0 "status=failed passes through" --claim build --status failed --evidence "$TMP/build-fail.log"
-expect_exit 0 "status=pending passes through" --claim build --status pending
-
-echo ""
-echo "→ 4. Custom success pattern override"
-printf 'CUSTOM_OK marker here\n' > "$TMP/custom.log"
-expect_exit 0 "custom success pattern matches" --claim test --status passed --evidence "$TMP/custom.log" --success-pattern 'CUSTOM_OK'
-expect_exit 1 "custom success pattern absent" --claim test --status passed --evidence "$TMP/build-ok.log" --success-pattern 'CUSTOM_OK'
-
-echo ""
-echo "→ 4b. Failure is decisive (closes the SUCCESS-substring bypass)"
-printf '** BUILD FAILED **\nnote: previous step SUCCESS (cached)\n' > "$TMP/bypass.log"
-expect_exit 1 "failing log that also says SUCCESS still fails" --claim build --status passed --evidence "$TMP/bypass.log"
-printf 'BUILD SUCCEEDED\nwarning: deprecated API used\n' > "$TMP/warn.log"
-expect_exit 0 "success log with a warning still passes" --claim build --status passed --evidence "$TMP/warn.log"
-
-echo ""
-echo "→ 4c. Bad/oversized custom patterns are a clean usage error (no crash)"
-expect_exit 2 "invalid regex pattern -> usage exit 2" --claim test --status passed --evidence "$TMP/test-ok.log" --success-pattern '('
-OVERLONG=$(printf 'a%.0s' $(seq 1 250))
-expect_exit 2 "overlong pattern -> usage exit 2" --claim test --status passed --evidence "$TMP/test-ok.log" --success-pattern "$OVERLONG"
-
-echo ""
-echo "→ 5. Wiring: phase docs reference the gate"
-GREP_HITS=0
-for doc in phase-3-dev phase-4-review phase-6-commit; do
-  grep -q "evidence-gate" "$ROOT/pipeline/multi-agent-refs/phases/$doc.md" && GREP_HITS=$((GREP_HITS+1))
-done
-if [ "$GREP_HITS" -ge 2 ]; then
-  pass "evidence-gate referenced in $GREP_HITS phase docs"
-else
-  fail "evidence-gate referenced in only $GREP_HITS phase docs (expected >= 2)"
-fi
-
-echo ""
-echo "══ evidence-gate smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]