@mmerterden/multi-agent-pipeline 12.7.0 → 12.9.0

package/pipeline/scripts/smoke-metrics-cache-ratio.sh

@@ -1,72 +0,0 @@
-#!/usr/bin/env bash
-# smoke-metrics-cache-ratio.sh  -  aggregate-metrics.mjs surfaces prompt-cache
-# reuse. The data path (log-metric.sh accepts tokens_cached) already exists; this
-# guards that the aggregator actually reads and reports it per model + overall,
-# in all three output modes (json / markdown / text).
-#
-# Exit 0 = all pass, 1 = any failure.
-
-set -uo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-AGG="$ROOT/pipeline/scripts/aggregate-metrics.mjs"
-PASS=0; FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-command -v node >/dev/null 2>&1 || { echo "error: node required" >&2; exit 127; }
-[ -f "$AGG" ] || { echo "FAIL: aggregate-metrics.mjs missing" >&2; exit 1; }
-
-cleanup() { [ -n "${TMP:-}" ] && rm -rf "$TMP"; }
-trap cleanup EXIT
-TMP=$(mktemp -d)
-
-FIX="$TMP/metrics.jsonl"
-# opus: 80k fresh input + 20k cached = cache_ratio 0.20
-# sonnet: 50k fresh input + 50k cached = cache_ratio 0.50
-# overall: 30k cached / (130k in + 30k... ) -> computed below by the script.
-{
-  echo '{"ts":"2026-06-19T10:00:00Z","task_id":"T-1","phase":"4","event":"review.completed","details":{"model":"opus","tokens_in":80000,"tokens_out":4000,"tokens_cached":20000}}'
-  echo '{"ts":"2026-06-19T10:01:00Z","task_id":"T-1","phase":"3","event":"dev.completed","details":{"model":"sonnet","tokens_in":50000,"tokens_out":6000,"tokens_cached":50000}}'
-} > "$FIX"
-
-echo "→ 1. JSON: per-model cache_ratio computed"
-out=$(METRICS_FILE="$FIX" node "$AGG" --json 2>/dev/null)
-check_json() { if node -e "const s=require('$TMP/out.json'); process.exit($1?0:1)" 2>/dev/null; then pass "$2"; else fail "$2"; fi; }
-printf '%s' "$out" > "$TMP/out.json"
-check_json 's.cost_per_model.opus.tokens_cached===20000' "opus cached tokens summed"
-check_json 's.cost_per_model.opus.cache_ratio===0.2' "opus cache_ratio = 0.20 (20k/100k)"
-check_json 's.cost_per_model.sonnet.cache_ratio===0.5' "sonnet cache_ratio = 0.50 (50k/100k)"
-
-echo ""
-echo "→ 2. JSON: overall cache_reuse block"
-check_json 's.cache_reuse.tokens_cached===70000' "overall cached = 70k"
-check_json 's.cache_reuse.tokens_in===130000' "overall fresh input = 130k"
-check_json 's.cache_reuse.cache_ratio===0.35' "overall cache_ratio = 0.35 (70k/200k)"
-
-echo ""
-echo "→ 3. markdown + text modes render cache column"
-md=$(METRICS_FILE="$FIX" node "$AGG" --markdown 2>/dev/null)
-case "$md" in *"Cache %"*) pass "markdown has Cache % column" ;; *) fail "markdown missing Cache % column" ;; esac
-case "$md" in *"prompt-cache reuse"*) pass "markdown has overall reuse line" ;; *) fail "markdown missing overall reuse line" ;; esac
-txt=$(METRICS_FILE="$FIX" node "$AGG" 2>/dev/null)
-case "$txt" in *"cache%"*) pass "text has cache% column" ;; *) fail "text missing cache% column" ;; esac
-case "$txt" in *"overall prompt-cache reuse"*) pass "text has overall reuse line" ;; *) fail "text missing overall reuse line" ;; esac
-
-echo ""
-echo "→ 4. no-cache data: ratio is 0% (input but no cache), null only when no input"
-NOFIX="$TMP/nocache.jsonl"
-{
-  # has input, zero cache -> ratio 0.00 (meaningful: 0% cache hit)
-  echo '{"ts":"2026-06-19T10:00:00Z","task_id":"T-2","phase":"4","event":"review.completed","details":{"model":"haiku","tokens_in":1000,"tokens_out":100}}'
-  # model tagged but no token counts at all -> denom 0 -> ratio null (not NaN)
-  echo '{"ts":"2026-06-19T10:00:01Z","task_id":"T-2","phase":"0","event":"init.completed","details":{"model":"none"}}'
-} > "$NOFIX"
-nout=$(METRICS_FILE="$NOFIX" node "$AGG" --json 2>/dev/null)
-printf '%s' "$nout" > "$TMP/out.json"
-check_json 's.cost_per_model.haiku.tokens_cached===0' "absent cached defaults to 0"
-check_json 's.cost_per_model.haiku.cache_ratio===0' "input-but-no-cache ratio is 0 (0% hit)"
-check_json 's.cost_per_model.none.cache_ratio===null' "no-input model ratio is null (not NaN)"
-
-echo ""
-echo "══ metrics cache-ratio smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]

package/pipeline/scripts/smoke-migrate-state.sh

@@ -1,102 +0,0 @@
-#!/usr/bin/env bash
-# smoke-migrate-state.sh  -  verify migrate-state.mjs behavior with synthetic
-# migrations. Does not depend on any real migration file existing  -  it lays
-# one down in a temp dir and points the runner at a fixture.
-
-set -uo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
-RUNNER="$SCRIPT_DIR/migrate-state.mjs"
-
-if [ ! -f "$RUNNER" ]; then
-  echo "FAIL: migrate-state.mjs not found" >&2
-  exit 1
-fi
-
-PASS=0
-FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-
-TMPDIR=$(mktemp -d) || { echo "FAIL: mktemp" >&2; exit 1; }
-trap 'rm -rf "$TMPDIR"' EXIT
-
-# Test 1  -  file at current version exits 0, no migration needed
-cat > "$TMPDIR/agent-state.json" <<'EOF'
-{"schemaVersion":"2.1.0","foo":"bar"}
-EOF
-OUT=$(node "$RUNNER" "$TMPDIR/agent-state.json" 2>&1)
-if echo "$OUT" | grep -q "already at schemaVersion 2.1.0"; then
-  pass "current-version file is a no-op"
-else
-  fail "current-version file should be a no-op (got: $OUT)"
-fi
-
-# Test 2  -  missing schemaVersion exits 1
-echo '{"foo":"bar"}' > "$TMPDIR/bad.json"
-node "$RUNNER" --type=state "$TMPDIR/bad.json" 2>/dev/null
-if [ $? -eq 1 ]; then
-  pass "missing schemaVersion exits 1"
-else
-  fail "missing schemaVersion did not exit 1"
-fi
-
-# Test 3  -  invalid JSON exits 1
-echo 'not-json' > "$TMPDIR/junk.json"
-node "$RUNNER" --type=state "$TMPDIR/junk.json" 2>/dev/null
-if [ $? -eq 1 ]; then
-  pass "invalid JSON exits 1"
-else
-  fail "invalid JSON did not exit 1"
-fi
-
-# Test 4  -  missing file exits 1
-node "$RUNNER" --type=state "$TMPDIR/does-not-exist.json" 2>/dev/null
-if [ $? -eq 1 ]; then
-  pass "missing file exits 1"
-else
-  fail "missing file did not exit 1"
-fi
-
-# Test 5  -  no migration path (older version than any migration)
-echo '{"schemaVersion":"0.1.0"}' > "$TMPDIR/ancient.json"
-node "$RUNNER" --type=state "$TMPDIR/ancient.json" 2>/dev/null
-if [ $? -eq 2 ]; then
-  pass "no migration path exits 2"
-else
-  fail "no migration path did not exit 2"
-fi
-
-# Test 6  -  unknown flag exits 1
-node "$RUNNER" --nonsense "$TMPDIR/agent-state.json" 2>/dev/null
-if [ $? -eq 1 ]; then
-  pass "unknown flag exits 1"
-else
-  fail "unknown flag did not exit 1"
-fi
-
-# Test 7  -  type auto-detection works (feed a doc already at the latest version)
-cat > "$TMPDIR/multi-agent-preferences.json" <<'EOF'
-{"schemaVersion":"2.4.0","global":{"identities":[],"keychainMapping":{},"recentAccounts":[],"accounts":[],"routines":[]},"projects":{}}
-EOF
-OUT=$(node "$RUNNER" "$TMPDIR/multi-agent-preferences.json" 2>&1)
-if echo "$OUT" | grep -q "already at schemaVersion 2.4.0"; then
-  pass "prefs type auto-detected from filename"
-else
-  fail "prefs auto-detection failed (got: $OUT)"
-fi
-
-# Test 8  -  ambiguous path requires --type
-cat > "$TMPDIR/ambig.json" <<'EOF'
-{"schemaVersion":"2.0.0"}
-EOF
-node "$RUNNER" "$TMPDIR/ambig.json" 2>/dev/null
-if [ $? -eq 1 ]; then
-  pass "ambiguous filename requires --type"
-else
-  fail "ambiguous filename did not require --type"
-fi
-
-echo ""
-echo "══ migrate-state smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ] || exit 1

package/pipeline/scripts/smoke-mode-dispatch-drift.sh

@@ -1,161 +0,0 @@
-#!/usr/bin/env bash
-#
-# smoke-mode-dispatch-drift.sh  -  assert each mode dispatch file's tracker
-# section is consistent with the canonical generator output.
-#
-# Each mode dispatch file (dev/SKILL.md, dev-autopilot/SKILL.md, dev-local/SKILL.md,
-# dev-local-autopilot/SKILL.md, autopilot/SKILL.md, local/SKILL.md, local-autopilot/SKILL.md) must
-# carry an inline "## Required: Phase Tracker Contract" section. v7.9.1 added
-# this contract; v8.0.0 added gen-mode-dispatch.mjs as the canonical
-# generator.
-#
-# As of v8.0.0 this smoke ENFORCES byte equality with the generator output.
-# Each mode dispatch file's tracker section must match `node gen-mode-dispatch.mjs
-# --mode=<name>` exactly. The generator is the source of truth; the on-disk
-# section is regenerated from it, never hand-edited.
-#
-# Mode → mode-name mapping (each pair is `<file>:<gen-mode>`):
-#   dev/SKILL.md                   → dev
-#   dev-autopilot/SKILL.md         → dev-autopilot
-#   dev-local/SKILL.md             → dev-local
-#   dev-local-autopilot/SKILL.md   → dev-local-autopilot
-#   autopilot/SKILL.md             → autopilot
-#   local/SKILL.md                 → local
-#   local-autopilot/SKILL.md       → local-autopilot
-#
-# Exit codes: 0=pass, 1=drift detected
-set -euo pipefail
-
-REPO_ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-CMD_DIR="$REPO_ROOT/pipeline/commands/multi-agent"
-GEN="$REPO_ROOT/pipeline/scripts/gen-mode-dispatch.mjs"
-
-[ -f "$GEN" ] || { echo "FAIL: gen-mode-dispatch.mjs missing"; exit 2; }
-
-declare -a MODE_FAST_WITH_TEST=("dev/SKILL.md")
-declare -a MODE_FAST_NO_TEST=("dev-autopilot/SKILL.md" "dev-local/SKILL.md" "dev-local-autopilot/SKILL.md")
-declare -a MODE_FAST=("${MODE_FAST_WITH_TEST[@]}" "${MODE_FAST_NO_TEST[@]}")
-declare -a MODE_FULL=("autopilot/SKILL.md" "local/SKILL.md" "local-autopilot/SKILL.md")
-
-PASS=0
-FAIL=0
-pass() { echo "  ✓ $*"; PASS=$((PASS + 1)); }
-fail() { echo "  ✗ $*"; FAIL=$((FAIL + 1)); }
-
-assert_section_present() {
-  local file="$1"
-  local count
-  count=$(grep -c '^## Required: Phase Tracker Contract' "$CMD_DIR/$file" || true)
-  case "$count" in
-    1) pass "$file declares tracker section exactly once" ;;
-    0) fail "$file is missing the tracker section" ;;
-    *) fail "$file has $count tracker sections (must be exactly 1)" ;;
-  esac
-}
-
-assert_phases() {
-  local file="$1"
-  shift
-  local expected_phases=("$@")
-  local section
-  section=$(awk '/^## Required: Phase Tracker Contract/{p=1} p{print} /^## /&&!/Required: Phase Tracker Contract/{p=0}' "$CMD_DIR/$file")
-  for ph in "${expected_phases[@]}"; do
-    if echo "$section" | grep -Fq "\"$ph\""; then
-      pass "$file declares phase $ph"
-    else
-      fail "$file missing phase $ph (expected for its mode family)"
-    fi
-  done
-}
-
-assert_ref_link() {
-  local file="$1"
-  if grep -Fq 'refs/tracker-contract.md' "$CMD_DIR/$file"; then
-    pass "$file links the tracker contract ref"
-  else
-    fail "$file missing tracker-contract.md link"
-  fi
-}
-
-echo "→ 1. Mode files declare the tracker contract section exactly once"
-for f in "${MODE_FAST[@]}" "${MODE_FULL[@]}"; do
-  assert_section_present "$f"
-done
-
-echo ""
-echo "→ 2a. Fast-mode files that keep User Test (dev/SKILL.md) declare 5 phases (0/3/5/6/7)"
-for f in "${MODE_FAST_WITH_TEST[@]}"; do
-  assert_phases "$f" "0:Init" "3:Dev" "5:Test" "6:Commit" "7:Report"
-done
-
-echo ""
-echo "→ 2b. Fast-mode files that skip User Test (dev-autopilot/local/local-autopilot) declare 4 phases (0/3/6/7)"
-for f in "${MODE_FAST_NO_TEST[@]}"; do
-  assert_phases "$f" "0:Init" "3:Dev" "6:Commit" "7:Report"
-done
-
-echo ""
-echo "→ 3. Autopilot/local full-pipeline files declare 7 phases (skip User Test)"
-# Phase 5 (User Test) needs an interactive prompt AND a worktree checkout, so it
-# runs only in the interactive worktree-backed modes (dev, full). Every autopilot
-# or local variant drops it.
-for f in "${MODE_FULL[@]}"; do
-  assert_phases "$f" "0:Init" "1:Analysis" "2:Planning" "3:Dev" "4:Review" "6:Commit" "7:Report"
-done
-
-echo ""
-echo "→ 4. Every mode file links refs/tracker-contract.md"
-for f in "${MODE_FAST[@]}" "${MODE_FULL[@]}"; do
-  assert_ref_link "$f"
-done
-
-echo ""
-echo "→ 5. Generator emits valid output for each mode"
-for mode in dev dev-autopilot dev-local dev-local-autopilot autopilot full local local-autopilot full-local; do
-  if node "$GEN" --mode="$mode" >/dev/null 2>&1; then
-    pass "gen-mode-dispatch.mjs --mode=$mode emits cleanly"
-  else
-    fail "gen-mode-dispatch.mjs --mode=$mode crashed or non-zero"
-  fi
-done
-
-if node "$GEN" --mode=bogus >/dev/null 2>&1; then
-  fail "gen-mode-dispatch.mjs accepted unknown mode 'bogus'  -  exit guard missing"
-else
-  pass "gen-mode-dispatch.mjs rejects unknown mode (exit 2)"
-fi
-
-echo ""
-echo "→ 6. Each mode file's tracker section is byte-equal to the generator output"
-# This is the strict gate: the generator is the source of truth, on-disk
-# sections are regenerated from it. Drift here means someone hand-edited a
-# mode file's tracker section without updating the generator. Fix:
-#   node pipeline/scripts/gen-mode-dispatch.mjs --mode=<m> > /tmp/gen.md
-#   <splice into the file's tracker section, or run the helper sync script>
-for pair in "dev/SKILL.md:dev" "dev-autopilot/SKILL.md:dev-autopilot" "dev-local/SKILL.md:dev-local" \
-            "dev-local-autopilot/SKILL.md:dev-local-autopilot" "autopilot/SKILL.md:autopilot" \
-            "local/SKILL.md:local" "local-autopilot/SKILL.md:local-autopilot"; do
-  file="${pair%:*}"
-  mode="${pair#*:}"
-  path="$CMD_DIR/$file"
-  expected=$(node "$GEN" --mode="$mode")
-  # Extract on-disk section: from "## Required: Phase Tracker Contract" up to
-  # (but not including) the next top-level "## " heading. Trim trailing blank
-  # lines so the generator (which ends with one trailing newline) matches.
-  actual=$(awk '
-    /^## Required: Phase Tracker Contract/ { p=1 }
-    p && /^## / && !/^## Required: Phase Tracker Contract/ { exit }
-    p { print }
-  ' "$path")
-  if [ "$(printf '%s' "$expected" | sha256sum | cut -d" " -f1)" = \
-       "$(printf '%s' "$actual"   | sha256sum | cut -d" " -f1)" ]; then
-    pass "$file tracker section matches gen-mode-dispatch.mjs --mode=$mode"
-  else
-    fail "$file tracker section drift vs --mode=$mode"
-    diff <(printf '%s\n' "$expected") <(printf '%s\n' "$actual") | head -20 | sed 's/^/      /'
-  fi
-done
-
-echo ""
-echo "══ smoke-mode-dispatch-drift: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -gt 0 ] && exit 1 || exit 0

package/pipeline/scripts/smoke-model-fallback.sh

@@ -1,89 +0,0 @@
-#!/usr/bin/env bash
-# smoke-model-fallback.sh  -  lint the model fallback contract wiring (v10.1.0).
-# The contract lets preferredModel (opus) personas degrade to sonnet on a
-# plan-window date gate, a dispatch error, or a budget ceiling - via
-# PHASE_MODEL_OVERRIDE, never by editing persona files. This smoke asserts the
-# doc, the prefs knob, and the phase wiring all exist and agree. No network,
-# no credentials.
-
-set -uo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
-REPO_ROOT="$(cd "$SCRIPT_DIR/../.." && pwd)"
-DOC="$REPO_ROOT/pipeline/multi-agent-refs/features/model-fallback.md"
-TEMPLATE="$REPO_ROOT/pipeline/preferences-template.json"
-P0="$REPO_ROOT/pipeline/multi-agent-refs/phases/phase-0-init.md"
-P4="$REPO_ROOT/pipeline/multi-agent-refs/phases/phase-4-review.md"
-
-PASS=0
-FAIL=0
-ok()   { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1" >&2; }
-
-echo "→ 1. Contract doc exists and covers the three triggers"
-if [ -f "$DOC" ]; then
-  ok "model-fallback.md present"
-  grep -q "premiumTierUntil" "$DOC" && ok "doc documents premiumTierUntil date gate" || fail "doc missing premiumTierUntil"
-  grep -q "onDispatchError" "$DOC" && ok "doc documents dispatch-error retry" || fail "doc missing onDispatchError"
-  grep -q "cost-budget-check" "$DOC" && ok "doc documents budget-ceiling trigger" || fail "doc missing budget trigger"
-  grep -q "PHASE_MODEL_OVERRIDE" "$DOC" && ok "doc uses the existing override mechanism" || fail "doc missing PHASE_MODEL_OVERRIDE"
-  grep -q "opus -> sonnet" "$DOC" && ok "tier ladder declared" || fail "tier ladder missing"
-else
-  fail "model-fallback.md missing"
-fi
-
-echo "→ 2. Prefs template carries the knob with safe defaults"
-if command -v python3 >/dev/null 2>&1; then
-  RESULT=$(python3 - "$TEMPLATE" <<'PYEOF'
-import json, sys
-t = json.load(open(sys.argv[1]))
-mf = t.get("global", {}).get("modelFallback")
-if not isinstance(mf, dict):
-    print("missing"); raise SystemExit
-checks = [
-    mf.get("enabled") is True,
-    "premiumTierUntil" in mf and mf["premiumTierUntil"] is None,
-    mf.get("fallbackModel") == "sonnet",
-    mf.get("onDispatchError") is True,
-]
-print("ok" if all(checks) else "bad-defaults")
-PYEOF
-)
-  case "$RESULT" in
-    ok) ok "template modelFallback knob present with safe defaults" ;;
-    missing) fail "template missing global.modelFallback" ;;
-    *) fail "template modelFallback defaults wrong ($RESULT)" ;;
-  esac
-else
-  echo "  (python3 unavailable - template check skipped)"
-fi
-
-echo "→ 3. Phase wiring references the contract"
-grep -q "model-fallback.md" "$P0" && ok "phase-0-init wires the date gate" || fail "phase-0-init missing fallback wiring"
-grep -q "model-fallback.md" "$P4" && ok "phase-4-review wires the per-dispatch triggers" || fail "phase-4-review missing fallback wiring"
-
-echo "→ 4. Top-tier personas declare their preferred model (fallback is dispatch-time, not file-time)"
-# Fable 5 is the top tier again: architect + Reviewer-1 (+ triage) personas pin it.
-FABLE_COUNT=$(grep -l "^preferredModel: fable" "$REPO_ROOT"/pipeline/agents/*.md 2>/dev/null | wc -l | tr -d ' ')
-if [ "$FABLE_COUNT" -ge 4 ]; then
-  ok "fable top-tier personas present ($FABLE_COUNT files)"
-else
-  fail "expected >=4 fable personas (architects + Reviewer-1), found $FABLE_COUNT"
-fi
-# Opus stays a valid tier - the first fallback below fable, and the top tier for security personas.
-OPUS_COUNT=$(grep -l "^preferredModel: opus" "$REPO_ROOT"/pipeline/agents/*.md 2>/dev/null | wc -l | tr -d ' ')
-if [ "$OPUS_COUNT" -ge 1 ]; then
-  ok "opus personas intact ($OPUS_COUNT files)"
-else
-  fail "expected >=1 opus persona (fallback tier / security), found $OPUS_COUNT"
-fi
-
-echo ""
-TOTAL=$((PASS+FAIL))
-if [ $FAIL -eq 0 ]; then
-  echo "══ model-fallback smoke: $PASS passed, 0 failed ══"
-  exit 0
-else
-  echo "══ model-fallback smoke: $PASS passed, $FAIL failed (of $TOTAL) ══" >&2
-  exit 1
-fi

package/pipeline/scripts/smoke-multi-repo-integration.sh

@@ -1,116 +0,0 @@
-#!/usr/bin/env bash
-# smoke-multi-repo-integration.sh  -  verify v5.6.0 multi-repo integration-build
-# contract: prefs schema accepts valid entries, rejects malformed ones, and the
-# contract doc ships with the required sections (lookup logic, learn-once
-# prompt, autopilot refuse-to-prompt, autopilot refuse-to-prompt behavior).
-#
-# This smoke validates STRUCTURE + CONTRACT. Integration-build execution logic
-# itself is LLM-orchestrated (the pipeline is the caller), so behavior tests
-# live in the integration-test tier, not smoke.
-
-set -uo pipefail
-
-HERE="$(cd "$(dirname "$0")" && pwd)"
-ROOT="$(cd "$HERE/../.." && pwd)"
-SCHEMA="$ROOT/pipeline/schemas/prefs.schema.json"
-DOC="$ROOT/pipeline/multi-agent-refs/multi-repo-integration-build.md"
-
-PASS=0
-FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-
-command -v jq >/dev/null || { echo "FAIL: jq required" >&2; exit 1; }
-
-echo "→ 1. prefs.schema.json defines multiRepoIntegrationHosts"
-jq -e '.properties.global.properties.multiRepoIntegrationHosts' "$SCHEMA" >/dev/null 2>&1 \
-  && pass "schema has global.multiRepoIntegrationHosts" \
-  || fail "schema missing multiRepoIntegrationHosts"
-
-echo ""
-echo "→ 2. schema item has required fields (repoSet) and proper constraints"
-SCHEMA_ITEM=$(jq '.properties.global.properties.multiRepoIntegrationHosts.items' "$SCHEMA")
-echo "$SCHEMA_ITEM" | jq -e '.required | index("repoSet")' >/dev/null 2>&1 \
-  && pass "item requires repoSet" || fail "item should require repoSet"
-echo "$SCHEMA_ITEM" | jq -e '.properties.repoSet.minItems == 2' >/dev/null 2>&1 \
-  && pass "repoSet minItems=2" || fail "repoSet should enforce minItems=2"
-echo "$SCHEMA_ITEM" | jq -e '.properties.platform.enum | index("ios") and index("android") and index("mixed")' >/dev/null 2>&1 \
-  && pass "platform enum covers ios/android/mixed" || fail "platform enum incomplete"
-echo "$SCHEMA_ITEM" | jq -e '.properties.noHost.type == "boolean"' >/dev/null 2>&1 \
-  && pass "noHost flag supports skip-combo" || fail "noHost flag missing"
-echo "$SCHEMA_ITEM" | jq -e '.properties.lastResult.enum | index("success") and index("failed") and index("skipped")' >/dev/null 2>&1 \
-  && pass "lastResult enum covers all terminal states" || fail "lastResult enum incomplete"
-
-echo ""
-echo "→ 3. contract doc exists and ships with required sections"
-[ -f "$DOC" ] && pass "multi-repo-integration-build.md exists" || { fail "doc missing"; echo "══ smoke: $PASS passed, $FAIL failed ══"; exit 1; }
-grep -q "## Why this exists" "$DOC" && pass "doc has 'Why this exists'" || fail "missing 'Why this exists'"
-grep -q "## When this rule fires" "$DOC" && pass "doc has 'When this rule fires'" || fail "missing trigger section"
-grep -q "Learn-once prompt" "$DOC" && pass "doc has learn-once prompt spec" || fail "missing learn-once prompt"
-grep -q "Autopilot behavior" "$DOC" && pass "doc has autopilot behavior" || fail "missing autopilot behavior"
-grep -q "noHost" "$DOC" && pass "doc documents noHost entries" || fail "missing noHost handling"
-grep -qE "phase-tracker\.sh sub 6 0" "$DOC" && pass "doc wires tracker sub-step" || fail "missing tracker wiring"
-
-echo ""
-echo "→ 4. phase-6-commit.md references the new Step 0"
-PHASE6="$ROOT/pipeline/multi-agent-refs/phases/phase-6-commit.md"
-grep -q "Step 0  -  Multi-Repo Integration Build" "$PHASE6" \
-  && pass "phase-6-commit has Step 0 entry" || fail "phase-6-commit missing Step 0"
-grep -q "multi-repo-integration-build.md" "$PHASE6" \
-  && pass "phase-6-commit links the contract doc" || fail "phase-6-commit missing contract link"
-
-echo ""
-echo "→ 5. generateCopilotInstructions includes the generic rule"
-# v8.0.0: install.js is a thin shim; the instruction body lives in
-# install/templates/copilot-instructions.md (Markdown source) and is read by
-# install/_copilot-instructions.mjs. The grep target is the source-of-truth
-# Markdown file.
-INSTALL_TPL="$ROOT/install/templates/copilot-instructions.md"
-[ -f "$INSTALL_TPL" ] || fail "install/templates/copilot-instructions.md missing  -  v8.0.0 generator template missing"
-grep -q "Post-Development Integration Build (Multi-Repo)" "$INSTALL_TPL" \
-  && pass "generator template includes generic rule heading" || fail "generator template missing generic rule"
-grep -q "multiRepoIntegrationHosts" "$INSTALL_TPL" \
-  && pass "generator template references registry" || fail "generator template missing registry reference"
-
-echo ""
-echo "→ 6. repoSet sorted-equality matching logic (spec check)"
-# Synthesize two entries, confirm match logic shape described by the doc.
-TMP=$(mktemp -d)
-trap 'rm -rf "$TMP"' EXIT
-cat > "$TMP/prefs.json" <<'EOF'
-{
-  "schemaVersion": "2.1.0",
-  "global": {
-    "multiRepoIntegrationHosts": [
-      { "repoSet": ["common-lib", "ui-components"], "hostPath": "/host/proj", "platform": "ios", "hostScheme": "UI", "count": 2, "lastResult": "success" },
-      { "repoSet": ["foo", "bar"], "noHost": true }
-    ]
-  },
-  "projects": {}
-}
-EOF
-
-# Lookup: sort incoming repoSet and compare to each entry's sorted repoSet.
-MATCH=$(jq -r --arg combo "common-lib,ui-components" \
-  '.global.multiRepoIntegrationHosts[]
-   | select((.repoSet | sort | join(",")) == $combo)
-   | .hostPath // "NOHOST"' "$TMP/prefs.json")
-[ "$MATCH" = "/host/proj" ] && pass "sorted-equality lookup returns hostPath" || fail "lookup returned: $MATCH"
-
-# Reversed input order still matches
-MATCH=$(jq -r --arg combo "common-lib,ui-components" \
-  '.global.multiRepoIntegrationHosts[]
-   | select((.repoSet | sort | join(",")) == $combo)
-   | .hostPath // "NOHOST"' "$TMP/prefs.json")
-[ "$MATCH" = "/host/proj" ] && pass "reversed-order input matches same host" || fail "reversed-order did not match"
-
-# noHost combo returns skip-intent
-NOHOST=$(jq -r --arg combo "bar,foo" \
-  '.global.multiRepoIntegrationHosts[]
-   | select((.repoSet | sort | join(",")) == $combo)
-   | .noHost // false' "$TMP/prefs.json")
-[ "$NOHOST" = "true" ] && pass "noHost entry returned for marked combo" || fail "noHost lookup failed: $NOHOST"
-
-echo ""
-echo "══ multi-repo-integration smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ] || exit 1

package/pipeline/scripts/smoke-multi-repo-worktree.sh

@@ -1,61 +0,0 @@
-#!/usr/bin/env bash
-# smoke-multi-repo-worktree.sh  -  asserts Phase 3 multi-repo dev contract is
-# documented correctly. Doc-shape smoke (no actual git ops).
-
-set -uo pipefail
-
-PHASE3="$HOME/.claude/multi-agent-refs/phases/phase-3-dev.md"
-
-PASS=0; FAIL=0
-pass() { PASS=$((PASS+1)); echo "  ✓ $1"; }
-fail() { FAIL=$((FAIL+1)); echo "  ✗ $1"; }
-
-if [ ! -f "$PHASE3" ]; then
-  fail "phase-3-dev.md missing"; echo "══ aborted ══"; exit 2
-fi
-
-echo "→ 1. v2.1.0+ Multi-Repo Mode subsection present"
-grep -q "v2.1.0+ Multi-Repo Mode" "$PHASE3" \
-  && pass "Phase 3 has v2.1.0+ Multi-Repo Mode subsection" \
-  || fail "Phase 3 missing v2.1.0+ Multi-Repo Mode subsection"
-
-echo "→ 2. Todo tagging contract"
-grep -q "\"repo\":" "$PHASE3" \
-  && pass "Todo schema includes \"repo\" tag" \
-  || fail "Todo schema missing \"repo\" tag"
-grep -q "Per-todo worktree switch" "$PHASE3" \
-  && pass "Per-todo worktree switch documented" \
-  || fail "Per-todo worktree switch missing"
-grep -q 'cd "\$WT_PATH"' "$PHASE3" \
-  && pass "Documents cd \$WT_PATH for build/test" \
-  || fail "cd \$WT_PATH missing from doc"
-
-echo "→ 3. Single-ticket status update"
-grep -q "exactly \*\*once per todo\*\*" "$PHASE3" \
-  && pass "Status update fires once per todo (not per repo per todo)" \
-  || fail "Single-ticket status-update rule missing"
-
-echo "→ 4. Build serialization"
-grep -q "BUILD_LOCK" "$PHASE3" \
-  && pass "Multi-repo respects BUILD_LOCK" \
-  || fail "BUILD_LOCK reference missing"
-grep -q "dependsOn\|dependency order" "$PHASE3" \
-  && pass "Cross-repo dependency order respected" \
-  || fail "Cross-repo dependency order not documented"
-
-echo "→ 5. Failure isolation"
-grep -q "buildStatus" "$PHASE3" \
-  && pass "Per-repo buildStatus persisted" \
-  || fail "Per-repo buildStatus missing"
-grep -q "do NOT proceed to Phase 4" "$PHASE3" \
-  && pass "Pipeline blocks Phase 4 if any repo unbuilt" \
-  || fail "Phase 4 block rule missing"
-
-echo "→ 6. Telemetry per-repo"
-grep -q 'log-metric.sh.*build.completed.*repo=' "$PHASE3" \
-  && pass "Per-repo build.completed telemetry documented" \
-  || fail "Per-repo telemetry missing"
-
-echo ""
-echo "══ multi-repo-worktree smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]