@mmerterden/multi-agent-pipeline 12.7.0 → 12.8.0

package/pipeline/scripts/smoke-phase4-triage.sh

@@ -1,229 +0,0 @@
-#!/usr/bin/env bash
-# smoke-phase4-triage.sh  -  contract check for Phase 4 triage output.
-#
-# Does NOT hit real review agents. Instead:
-#   1. Synthesizes representative triage-output fixtures (valid + each edge case).
-#   2. Validates them against pipeline/schemas/triage-output.schema.json
-#      using a minimal JSON-structure check (zero-dep  -  same style as
-#      validate-schemas.mjs).
-#   3. Confirms the pipeline's documented edge-case rules hold for:
-#      - no-findings short-circuit (approved: true)
-#      - over-rejection guard threshold (>0.80 rejected)
-#      - schema rejects invented reviewer (e.g. "haiku"  -  removed in v2.1.0)
-#      - accepted finding without `fix` is rejected (fix is required on accepted)
-
-set -euo pipefail
-
-HERE="$(cd "$(dirname "$0")" && pwd)"
-ROOT="$(cd "$HERE/../.." && pwd)"
-SCHEMA="$ROOT/pipeline/schemas/triage-output.schema.json"
-TMP="$(mktemp -d -t multi-agent-smoke-XXXX)"
-trap 'rm -rf "$TMP"' EXIT
-
-PASS=0
-FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-
-need() { command -v "$1" >/dev/null 2>&1 || { echo "error: $1 is required" >&2; exit 127; }; }
-need node
-need jq
-
-[ -f "$SCHEMA" ] || { echo "error: schema not found at $SCHEMA" >&2; exit 1; }
-
-# Helper: delegate to the real runtime validator so this smoke test exercises
-# the same code path the pipeline uses in production. Exit codes:
-#   0 = valid and clean
-#   1 = structural / field violation
-#   2 = over-rejection guard tripped
-#   3 = contradiction auto-corrected
-check_via_validator() {
-  local file="$1"
-  local expect_exit="$2" # numeric exit code
-  local reason="$3"
-  local actual_exit=0
-  node "$ROOT/pipeline/scripts/validate-triage.mjs" "$file" >/tmp/validator-out 2>/tmp/validator-err || actual_exit=$?
-  if [ "$actual_exit" = "$expect_exit" ]; then
-    pass "$reason (exit=$actual_exit)"
-  else
-    fail "$reason (expected exit=$expect_exit, got $actual_exit; stderr: $(cat /tmp/validator-err))"
-  fi
-}
-
-echo "→ fixture 1: empty findings (short-circuit)"
-cat > "$TMP/empty.json" <<'EOF'
-{"accepted": [], "deferred": [], "rejected": [], "approved": true}
-EOF
-check_via_validator "$TMP/empty.json" 0 "empty triage output is valid (clean)"
-
-echo ""
-echo "→ fixture 2: one accepted blocking finding"
-cat > "$TMP/accepted.json" <<'EOF'
-{
-  "accepted": [
-    {"severity":"blocking","file":"src/auth.ts","line":42,"issue":"SQL injection via untrusted input","fix":"parameterize the query with prepared statement","reviewer":"opus"}
-  ],
-  "deferred": [],
-  "rejected": [],
-  "approved": false
-}
-EOF
-check_via_validator "$TMP/accepted.json" 0 "accepted blocking with fix + reviewer is valid"
-
-echo ""
-echo "→ fixture 3: accepted without fix (must reject  -  exit 1)"
-cat > "$TMP/no-fix.json" <<'EOF'
-{
-  "accepted": [
-    {"severity":"blocking","file":"src/auth.ts","line":42,"issue":"SQL injection","reviewer":"opus"}
-  ],
-  "deferred": [],
-  "rejected": [],
-  "approved": false
-}
-EOF
-check_via_validator "$TMP/no-fix.json" 1 "accepted without fix triggers exit 1"
-
-echo ""
-echo "→ fixture 4: reviewer=haiku (removed in v2.1.0, must reject  -  exit 1)"
-cat > "$TMP/haiku.json" <<'EOF'
-{
-  "accepted": [
-    {"severity":"suggestion","file":"x.ts","line":1,"issue":"naming","fix":"rename","reviewer":"haiku"}
-  ],
-  "deferred": [],
-  "rejected": [],
-  "approved": true
-}
-EOF
-check_via_validator "$TMP/haiku.json" 1 "haiku reviewer triggers exit 1"
-
-echo ""
-echo "→ fixture 5: over-rejection guard (5 rejected, 0 others  -  exit 2)"
-cat > "$TMP/over-reject.json" <<'EOF'
-{
-  "accepted": [],
-  "deferred": [],
-  "rejected": [
-    {"finding":{"severity":"blocking","file":"a.ts","line":1,"issue":"i111"},"reason":"false positive"},
-    {"finding":{"severity":"blocking","file":"b.ts","line":2,"issue":"i222"},"reason":"false positive"},
-    {"finding":{"severity":"important","file":"c.ts","line":3,"issue":"i333"},"reason":"duplicate"},
-    {"finding":{"severity":"blocking","file":"d.ts","line":4,"issue":"i444"},"reason":"style only"},
-    {"finding":{"severity":"suggestion","file":"e.ts","line":5,"issue":"i555"},"reason":"already correct"}
-  ],
-  "approved": true
-}
-EOF
-check_via_validator "$TMP/over-reject.json" 2 "100% rejection triggers over-rejection guard (exit 2)"
-
-echo ""
-echo "→ fixture 6: contradiction (approved=false but no blocker  -  exit 3 with corrected)"
-cat > "$TMP/contradiction.json" <<'EOF'
-{
-  "accepted": [
-    {"severity":"suggestion","file":"x.ts","line":1,"issue":"name change","fix":"rename to FooBar","reviewer":"opus"}
-  ],
-  "deferred": [],
-  "rejected": [],
-  "approved": false
-}
-EOF
-check_via_validator "$TMP/contradiction.json" 3 "contradiction auto-correct (exit 3)"
-# The validator exits 3 (non-zero) but we still want its stdout.
-CORRECTED_APPROVED=$(node "$ROOT/pipeline/scripts/validate-triage.mjs" "$TMP/contradiction.json" 2>/dev/null || true)
-if echo "$CORRECTED_APPROVED" | jq -e '.corrected.approved == true' >/dev/null 2>&1; then
-  pass "corrected.approved == true"
-else
-  fail "corrected.approved was not true"
-fi
-
-echo ""
-echo "→ fixture 7: optional consensus block, unverified verdict (valid  -  exit 0)"
-cat > "$TMP/consensus-unverified.json" <<'EOF'
-{
-  "accepted": [],
-  "deferred": [],
-  "rejected": [],
-  "approved": true,
-  "consensus": {
-    "reviewerCount": 2,
-    "verdict": "unverified",
-    "disagreements": [
-      {"file":"Auth/KeychainStore.swift","line":40,"issue":"token persisted without access-control flag","note":"both approved a keychain change  -  agreement unverified, confirm manually"}
-    ]
-  }
-}
-EOF
-check_via_validator "$TMP/consensus-unverified.json" 0 "optional consensus (unverified) is valid"
-
-echo ""
-echo "→ fixture 8: consensus split verdict (valid  -  exit 0)"
-cat > "$TMP/consensus-split.json" <<'EOF'
-{
-  "accepted": [
-    {"severity":"blocking","file":"src/pay.ts","line":7,"issue":"amount parsed as float loses precision","fix":"use integer minor units","reviewer":"opus"}
-  ],
-  "deferred": [],
-  "rejected": [],
-  "approved": false,
-  "consensus": {
-    "reviewerCount": 2,
-    "verdict": "split",
-    "disagreements": [
-      {"file":"src/pay.ts","line":7,"issue":"float money parsing","note":"Opus blocking, Sonnet approved"}
-    ]
-  }
-}
-EOF
-check_via_validator "$TMP/consensus-split.json" 0 "consensus split verdict is valid"
-
-echo ""
-echo "→ fixture 9: bad consensus verdict enum (must reject  -  exit 1)"
-cat > "$TMP/consensus-bad.json" <<'EOF'
-{
-  "accepted": [],
-  "deferred": [],
-  "rejected": [],
-  "approved": true,
-  "consensus": {"reviewerCount": 2, "verdict": "maybe"}
-}
-EOF
-check_via_validator "$TMP/consensus-bad.json" 1 "bad consensus verdict triggers exit 1"
-
-echo ""
-echo "→ fixture 10: consensus missing required reviewerCount (must reject  -  exit 1)"
-cat > "$TMP/consensus-noreqcount.json" <<'EOF'
-{
-  "accepted": [],
-  "deferred": [],
-  "rejected": [],
-  "approved": true,
-  "consensus": {"verdict": "unanimous-pass"}
-}
-EOF
-check_via_validator "$TMP/consensus-noreqcount.json" 1 "consensus without reviewerCount triggers exit 1"
-
-echo ""
-echo "→ fixture 11: consensus unanimous-block contradicts approved=true (exit 1)"
-cat > "$TMP/consensus-contra1.json" <<'EOF'
-{"accepted":[],"deferred":[],"rejected":[],"approved":true,"consensus":{"reviewerCount":2,"verdict":"unanimous-block"}}
-EOF
-check_via_validator "$TMP/consensus-contra1.json" 1 "unanimous-block + approved=true is rejected"
-
-echo ""
-echo "→ fixture 12: unanimous-pass contradicts an accepted blocker (exit 1)"
-cat > "$TMP/consensus-contra2.json" <<'EOF'
-{"accepted":[{"severity":"blocking","file":"a.ts","line":1,"issue":"sql injection here","fix":"parameterize the query","reviewer":"opus"}],"deferred":[],"rejected":[],"approved":false,"consensus":{"reviewerCount":2,"verdict":"unanimous-pass"}}
-EOF
-check_via_validator "$TMP/consensus-contra2.json" 1 "unanimous-pass + accepted blocker is rejected"
-
-echo ""
-echo "→ fixture 13: single reviewer cannot be unanimous (exit 1)"
-cat > "$TMP/consensus-solo.json" <<'EOF'
-{"accepted":[],"deferred":[],"rejected":[],"approved":true,"consensus":{"reviewerCount":1,"verdict":"unanimous-pass"}}
-EOF
-check_via_validator "$TMP/consensus-solo.json" 1 "reviewerCount 1 + unanimous verdict is rejected"
-
-echo ""
-echo "══ phase4-triage smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]

package/pipeline/scripts/smoke-plan-approval-gate.sh

@@ -1,71 +0,0 @@
-#!/usr/bin/env bash
-# smoke-plan-approval-gate.sh  -  contract check for Phase 2 Plan Approval Gate (v5.3.0).
-#
-# Verifies:
-#   1. phase-2-planning.md declares Step 5 as the gate (clarification + approval)
-#   2. Gate is explicitly scoped to non-autopilot, non-dev modes (autopilot + --dev skip)
-#   3. Clarification cap is 2 rounds
-#   4. agent-state schema exposes the six audit fields in phases.*
-#   5. Mode matrix in modes.md, autopilot.md, dev.md, dev-autopilot.md mention the gate
-#   6. SKILL.md autopilot matrix shows "Gate skip"
-
-set -uo pipefail
-
-HERE="$(cd "$(dirname "$0")" && pwd)"
-ROOT="$(cd "$HERE/../.." && pwd)"
-
-PASS=0
-FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-
-PHASE2="$ROOT/pipeline/multi-agent-refs/phases/phase-2-planning.md"
-SCHEMA="$ROOT/pipeline/schemas/agent-state.schema.json"
-SKILL="$ROOT/pipeline/skills/shared/core/multi-agent/SKILL.md"
-MODES="$ROOT/pipeline/multi-agent-refs/phases/modes.md"
-AUTOPILOT="$ROOT/pipeline/commands/multi-agent/autopilot/SKILL.md"
-DEV="$ROOT/pipeline/commands/multi-agent/dev/SKILL.md"
-DEVAUTO="$ROOT/pipeline/commands/multi-agent/dev-autopilot/SKILL.md"
-
-echo "→ 1. Phase 2 doc declares the gate + sub-modes"
-grep -q "Plan Approval Gate" "$PHASE2" && pass "phase-2 has 'Plan Approval Gate'" || fail "phase-2 missing gate heading"
-grep -q "Clarification Mode" "$PHASE2" && pass "phase-2 has Clarification Mode (5a)" || fail "phase-2 missing 5a"
-grep -q "Approval Loop" "$PHASE2" && pass "phase-2 has Approval Loop (5b)" || fail "phase-2 missing 5b"
-
-echo ""
-echo "→ 2. Gate is scoped  -  autopilot + --dev skip"
-grep -q "state.autopilot === true" "$PHASE2" && pass "phase-2 guards against autopilot" || fail "phase-2 missing autopilot guard"
-grep -q "state.onlyDevelop === true" "$PHASE2" && pass "phase-2 guards against --dev" || fail "phase-2 missing --dev guard"
-
-echo ""
-echo "→ 3. Clarification cap = 2 rounds"
-grep -qi "max 2 rounds" "$PHASE2" && pass "phase-2 states 'max 2 rounds'" || fail "phase-2 missing '2 rounds' cap"
-grep -q '"clarificationRounds"' "$SCHEMA" && pass "schema exposes clarificationRounds" || fail "schema missing clarificationRounds"
-grep -A4 '"clarificationRounds"' "$SCHEMA" | grep -q '"maximum": 2' && pass "clarificationRounds max=2" || fail "clarificationRounds schema cap wrong"
-
-echo ""
-echo "→ 4. Schema exposes all six audit fields"
-for field in clarificationRounds clarificationQuestions clarificationAnswers planIterations planApprovedAt planEditRequests; do
-  grep -q "\"$field\"" "$SCHEMA" && pass "schema has $field" || fail "schema missing $field"
-done
-
-echo ""
-echo "→ 5. Mode matrices reference the gate"
-grep -qi "Plan Approval Gate" "$MODES" && pass "modes.md references gate" || fail "modes.md missing gate"
-grep -qi "Gate skip\|Plan Approval Gate" "$AUTOPILOT" && pass "autopilot.md references gate skip" || fail "autopilot.md missing gate mention"
-grep -qi "Plan Approval Gate\|plan gate\|Plan Approval" "$DEV" && pass "dev.md references gate (skipped)" || fail "dev.md missing gate mention"
-grep -qi "Plan Approval Gate" "$DEVAUTO" && pass "dev-autopilot.md references gate (skipped)" || fail "dev-autopilot.md missing gate mention"
-
-echo ""
-echo "→ 6. SKILL.md autopilot matrix shows Gate skip semantics"
-grep -qi "Plan Approval Gate" "$SKILL" && pass "SKILL.md references gate" || fail "SKILL.md missing gate"
-grep -qi "Gate skip" "$SKILL" && pass "SKILL.md autopilot matrix shows 'Gate skip'" || fail "SKILL.md autopilot matrix missing 'Gate skip'"
-
-echo ""
-echo "→ 7. Free-text edit contract (no sentinel required)"
-grep -qi "free-text" "$PHASE2" && pass "phase-2 documents the free-text edit path" || fail "phase-2 missing free-text edit path"
-grep -q 'planEditRequests' "$PHASE2" && pass "phase-2 references planEditRequests" || fail "phase-2 missing planEditRequests"
-
-echo ""
-echo "══ plan-approval-gate smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]

package/pipeline/scripts/smoke-plan-safety.sh

@@ -1,139 +0,0 @@
-#!/usr/bin/env bash
-# smoke-plan-safety.sh  -  v7.0.G
-#
-# Verifies the Phase 2 autopilot safety classifier:
-#   1. Low-risk plan: score low, recommendPause=false
-#   2. High-risk plan (destructive + security + migration): recommendPause=true
-#   3. File-count thresholds fire: medium (8-15) vs high (>15)
-#   4. Security-path detection on /auth/, /keychain/, /credentials/, etc.
-#   5. delete-without-test rule fires when destructive task has no sibling test
-#   6. Malformed JSON input → exit 1
-#   7. Schema pref exposes autopilotSafetyGate default true
-#
-# Exit 0 all pass, 1 any failure.
-
-set -euo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-CLS="$ROOT/pipeline/scripts/classify-plan-safety.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' "$1"; }
-
-printf '→ smoke-plan-safety (v7.0.G): autopilot safety classifier contract\n'
-
-# --- 1: low-risk ---
-out=$(cat <<'JSON' | node "$CLS" -
-{"tasks":[{"id":"T1","title":"Rename a test helper","type":"test","files":["tests/helpers/fixtures.ts"]}],"approved":true}
-JSON
-)
-if [ "$(jq -r .recommendPause <<< "$out")" = "false" ]; then
-  record_pass "low-risk plan does not recommend pause"
-else
-  record_fail "low-risk plan should not recommend pause (got: $out)"
-fi
-
-# --- 2: high-risk (multi-signal) ---
-out=$(cat <<'JSON' | node "$CLS" -
-{"approach":"Drop auth schema and remove legacy keychain storage","tasks":[
-  {"id":"T1","title":"Drop auth_tokens migration","type":"code","files":["schemas/migrations/008.sql"]},
-  {"id":"T2","title":"Remove KeychainStore","type":"refactor","files":["Sources/Auth/KeychainStore.swift","Sources/Security/TokenManager.swift"]}
-],"approved":true}
-JSON
-)
-if [ "$(jq -r .recommendPause <<< "$out")" = "true" ]; then
-  record_pass "high-risk plan recommends pause"
-else
-  record_fail "high-risk plan should recommend pause (got: $out)"
-fi
-
-# --- 3: file-count thresholds ---
-files_json=$(jq -cn '[range(0;10)] | map("src/m\(.).kt")')
-out=$(cat <<JSON | node "$CLS" -
-{"tasks":[{"id":"T1","title":"touch 10 files","type":"code","files":$files_json}],"approved":true}
-JSON
-)
-rules=$(jq -r '.reasons[].rule' <<< "$out" | tr '\n' ' ')
-if echo "$rules" | grep -q 'file-count-medium'; then
-  record_pass "file-count-medium fires at 10 files"
-else
-  record_fail "file-count-medium should fire at 10 files (got rules: $rules)"
-fi
-
-files_json=$(jq -cn '[range(0;20)] | map("src/m\(.).kt")')
-out=$(cat <<JSON | node "$CLS" -
-{"tasks":[{"id":"T1","title":"touch 20 files","type":"code","files":$files_json}],"approved":true}
-JSON
-)
-rules=$(jq -r '.reasons[].rule' <<< "$out" | tr '\n' ' ')
-if echo "$rules" | grep -q 'file-count-high'; then
-  record_pass "file-count-high fires at 20 files"
-else
-  record_fail "file-count-high should fire at 20 files (got rules: $rules)"
-fi
-
-# --- 4: security-path ---
-out=$(cat <<'JSON' | node "$CLS" -
-{"tasks":[{"id":"T1","title":"add test","type":"test","files":["tests/keychain/helper.ts"]}],"approved":true}
-JSON
-)
-if jq -e '.reasons[] | select(.rule=="security-path")' <<< "$out" >/dev/null; then
-  record_pass "security-path detects /keychain/"
-else
-  record_fail "security-path should detect /keychain/"
-fi
-
-# --- 5: delete-without-test rule ---
-out=$(cat <<'JSON' | node "$CLS" -
-{"tasks":[
-  {"id":"T1","title":"Remove old feature flag handler","type":"refactor","files":["Sources/FeatureFlags/OldHandler.swift"]}
-],"approved":true}
-JSON
-)
-if jq -e '.reasons[] | select(.rule=="delete-without-test")' <<< "$out" >/dev/null; then
-  record_pass "delete-without-test fires when destructive task has no paired test task"
-else
-  record_fail "delete-without-test should fire here"
-fi
-
-# --- delete WITH paired test should NOT fire ---
-out=$(cat <<'JSON' | node "$CLS" -
-{"tasks":[
-  {"id":"T1","title":"Add regression test for old flag removal","type":"test","files":["Sources/FeatureFlags/OldHandlerTests.swift"]},
-  {"id":"T2","title":"Remove old feature flag handler","type":"refactor","files":["Sources/FeatureFlags/OldHandler.swift"]}
-],"approved":true}
-JSON
-)
-if ! jq -e '.reasons[] | select(.rule=="delete-without-test")' <<< "$out" >/dev/null; then
-  record_pass "delete-without-test does NOT fire when paired test exists"
-else
-  record_fail "delete-without-test should NOT fire when paired test exists"
-fi
-
-# --- 6: malformed input ---
-set +e
-echo 'not json' | node "$CLS" - >/dev/null 2>&1
-rc=$?
-set -e
-if [ "$rc" = "1" ]; then
-  record_pass "exit 1 on malformed JSON"
-else
-  record_fail "expected exit 1 on malformed JSON (got $rc)"
-fi
-
-# --- 7: schema pref ---
-if jq -e '.properties.global.properties.autopilotSafetyGate.default == true' "$SCHEMA" >/dev/null 2>&1; then
-  record_pass "prefs schema exposes autopilotSafetyGate with default true"
-else
-  record_fail "prefs schema missing autopilotSafetyGate or wrong default"
-fi
-
-printf '\n══ plan-safety 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-plan-todos.sh

@@ -1,196 +0,0 @@
-#!/usr/bin/env bash
-# smoke-plan-todos.sh
-#
-# Verifies the Plan-as-Todo surface:
-#   1. pipeline/lib/plan-todos.sh exists, parses cleanly
-#   2. Helper accepts: init, set, start, complete, fail, skip, next, list, status, show
-#   3. Unknown subcommand exits 64
-#   4. Schema pipeline/schemas/plan-todos.schema.json valid + required fields
-#   5. Lifecycle test on a temp state dir: set → next → start → complete → status
-#   6. next() respects deps (t2 deps on t1; with t1 pending, next returns t1; with t1 in_progress, next is empty)
-#   7. start() refuses when deps unsatisfied
-#   8. list() output uses [x]/[~]/[ ] markers
-#   9. prefs schema exposes planTodos.enabled (default false)
-#  10. Phase 2 doc documents Step 4.5
-#  11. Phase 3 doc documents plan-todos.sh iteration
-#  12. Phase 7 doc references Plan Todo Rollup
-#
-# Exit 0 = all pass, 1 = any failure.
-
-set -uo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-HELPER="$ROOT/pipeline/lib/plan-todos.sh"
-SCHEMA="$ROOT/pipeline/schemas/plan-todos.schema.json"
-PREFS="$ROOT/pipeline/schemas/prefs.schema.json"
-P2="$ROOT/pipeline/multi-agent-refs/phases/phase-2-planning.md"
-P3="$ROOT/pipeline/multi-agent-refs/phases/phase-3-dev.md"
-P7="$ROOT/pipeline/multi-agent-refs/phases/phase-7-report.md"
-
-pass=0
-fail=0
-failures=()
-record_pass() { pass=$((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' "$1"; }
-
-printf '→ smoke-plan-todos: Plan-as-Todo live list contract\n'
-
-# 1. Helper exists + parses
-if [ ! -f "$HELPER" ]; then
-  record_fail "pipeline/lib/plan-todos.sh missing"
-elif ! bash -n "$HELPER" 2>/dev/null; then
-  record_fail "plan-todos.sh syntax error"
-else
-  record_pass "plan-todos.sh exists and parses"
-fi
-
-# 2. Subcommands recognized (each prints either usage or runs)
-SUBS=(init set start complete fail skip next list status show)
-for s in "${SUBS[@]}"; do
-  if bash "$HELPER" "$s" --probe-noop 2>/dev/null >/dev/null; then
-    record_pass "subcommand '$s' accepted (probe path)"
-  elif bash "$HELPER" "$s" 2>&1 | grep -qE "usage: $s"; then
-    record_pass "subcommand '$s' prints usage"
-  else
-    # Some subcommands fail without args via exit 64  -  that also confirms they're known
-    bash "$HELPER" "$s" 2>/dev/null
-    rc=$?
-    if [ "$rc" = "64" ] || [ "$rc" = "1" ]; then
-      record_pass "subcommand '$s' recognized"
-    else
-      record_fail "subcommand '$s' not recognized (exit $rc)"
-    fi
-  fi
-done
-
-# 3. Unknown subcommand exits 64
-if bash "$HELPER" garbage 2>/dev/null >/dev/null; then
-  record_fail "unknown subcommand should exit non-zero"
-else
-  record_pass "unknown subcommand rejected"
-fi
-
-# 4. Schema valid
-if [ ! -f "$SCHEMA" ]; then
-  record_fail "plan-todos.schema.json missing"
-elif ! jq empty "$SCHEMA" 2>/dev/null; then
-  record_fail "plan-todos.schema.json is not valid JSON"
-else
-  record_pass "plan-todos.schema.json parses"
-  for prop in title todos; do
-    if jq -e ".properties.${prop}" "$SCHEMA" >/dev/null 2>&1; then
-      record_pass "schema requires '${prop}'"
-    else
-      record_fail "schema missing '${prop}'"
-    fi
-  done
-  if jq -e '.properties.todos.items.properties.status.enum
-            | (index("pending") != null and index("in_progress") != null
-               and index("completed") != null and index("skipped") != null
-               and index("failed") != null)' "$SCHEMA" >/dev/null 2>&1; then
-    record_pass "schema enforces 5-state status enum"
-  else
-    record_fail "schema status enum incomplete"
-  fi
-fi
-
-# 5-8. Lifecycle test against a temp state dir
-# Use a temp directory derived from $HOME so the path stays cross-user.
-TEST_TASK="SMOKE-PT-$$"
-TEST_PROJECT_KEY=$(printf '%s' "$HOME/test-plan-todos-smoke" | tr '/' '-')
-TEST_PROJECT_ROOT="$HOME/.claude/projects/$TEST_PROJECT_KEY"
-TEST_DIR="$TEST_PROJECT_ROOT/state/$TEST_TASK"
-mkdir -p "$TEST_DIR"
-echo '{"task":{"id":"'"$TEST_TASK"'"}}' > "$TEST_DIR/agent-state.json"
-PLAN='{"title":"Smoke","todos":[
-  {"id":"t1","task":"first","status":"pending","deps":[]},
-  {"id":"t2","task":"second","status":"pending","deps":["t1"]}
-]}'
-if bash "$HELPER" set "$TEST_TASK" "$PLAN" 2>/dev/null; then
-  record_pass "set writes plan blob"
-else
-  record_fail "set failed on fixture"
-fi
-
-next1=$(bash "$HELPER" next "$TEST_TASK" 2>/dev/null | cut -f1)
-if [ "$next1" = "t1" ]; then
-  record_pass "next returns t1 (deps empty)"
-else
-  record_fail "next should return t1, got '$next1'"
-fi
-
-bash "$HELPER" start "$TEST_TASK" t1 2>/dev/null
-
-next2=$(bash "$HELPER" next "$TEST_TASK" 2>/dev/null | cut -f1)
-if [ -z "$next2" ]; then
-  record_pass "next is empty while t1 is in_progress (deps not yet satisfied for t2)"
-else
-  record_fail "next should be empty during t1 in_progress, got '$next2'"
-fi
-
-# start t2 should refuse (deps not satisfied)
-if bash "$HELPER" start "$TEST_TASK" t2 2>/dev/null; then
-  record_fail "start should refuse t2 while t1 not completed"
-else
-  record_pass "start refuses t2 when t1 dep unsatisfied"
-fi
-
-bash "$HELPER" complete "$TEST_TASK" t1 "smoke note" 2>/dev/null
-
-next3=$(bash "$HELPER" next "$TEST_TASK" 2>/dev/null | cut -f1)
-if [ "$next3" = "t2" ]; then
-  record_pass "next returns t2 after t1 completed"
-else
-  record_fail "next should return t2 after t1 completed, got '$next3'"
-fi
-
-LIST_OUT=$(bash "$HELPER" list "$TEST_TASK" 2>/dev/null)
-if echo "$LIST_OUT" | grep -qE '\[x\].*t1' \
-   && echo "$LIST_OUT" | grep -qE '\[ \].*t2'; then
-  record_pass "list renders [x] for completed and [ ] for pending"
-else
-  record_fail "list markdown markers incorrect"
-fi
-
-STATUS_OUT=$(bash "$HELPER" status "$TEST_TASK" 2>/dev/null)
-if echo "$STATUS_OUT" | grep -qE '1/2 done'; then
-  record_pass "status reports '1/2 done'"
-else
-  record_fail "status should report '1/2 done', got '$STATUS_OUT'"
-fi
-
-# Cleanup the test state
-rm -rf "$TEST_PROJECT_ROOT"
-
-# 9. Prefs schema toggle
-if jq -e '.properties.global.properties.planTodos.properties.enabled
-          | has("default") and .default == false' "$PREFS" >/dev/null 2>&1; then
-  record_pass "prefs schema exposes planTodos.enabled (default false)"
-else
-  record_fail "prefs schema missing planTodos.enabled or wrong default"
-fi
-
-# 10-12. Phase docs wiring
-if grep -qE 'Step 4\.5  -  Emit Plan Todo List' "$P2"; then
-  record_pass "phase-2-planning.md documents Step 4.5"
-else
-  record_fail "phase-2-planning.md missing Step 4.5"
-fi
-if grep -qE 'plan-todos\.sh"? next' "$P3"; then
-  record_pass "phase-3-dev.md documents plan-todos.sh iteration"
-else
-  record_fail "phase-3-dev.md missing plan-todos.sh wiring"
-fi
-if grep -qE 'Plan Todo Rollup' "$P7"; then
-  record_pass "phase-7-report.md references Plan Todo Rollup"
-else
-  record_fail "phase-7-report.md missing Plan Todo Rollup section"
-fi
-
-printf '\n══ plan-todos smoke: %d passed, %d failed ══\n' "$pass" "$fail"
-if [ "$fail" -gt 0 ]; then
-  printf '\nFailures:\n'
-  for msg in "${failures[@]}"; do printf '  - %s\n' "$msg"; done
-  exit 1
-fi
-exit 0