@mmerterden/multi-agent-pipeline 12.6.0 → 12.8.0

package/pipeline/scripts/smoke-cost-summary.sh

@@ -1,139 +0,0 @@
-#!/usr/bin/env bash
-# smoke-cost-summary.sh  -  v6.1.J
-#
-# Verifies the cost-summary pipeline:
-#   1. cost-table.json parses and has opus/sonnet/haiku entries
-#   2. render-cost-summary.sh exits 2 when no tracker data available
-#   3. render-cost-summary.sh emits a valid markdown table from tracker fixture
-#   4. OTel-spans fallback path renders when tracker JSON is missing
-#   5. Unknown model gracefully renders USD as  -  with footnote
-#   6. Schema (prefs.schema.json) exposes costSummary toggle
-#   7. channels.md documents Cost Summary content option
-#
-# Exit 0 = all pass, 1 = any failure.
-
-set -euo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-RENDERER="$ROOT/pipeline/scripts/render-cost-summary.sh"
-TABLE="$ROOT/pipeline/scripts/cost-table.json"
-SCHEMA="$ROOT/pipeline/schemas/prefs.schema.json"
-CHANNELS_DOC="$ROOT/pipeline/commands/multi-agent/channels/SKILL.md"
-
-pass=0
-fail=0
-failures=()
-record_pass() { pass=$((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' "$1"; }
-
-printf '→ smoke-cost-summary (v6.1.J): PR-body Cost Summary renderer contract\n'
-
-# --- 1: cost-table.json shape ---
-if jq -e '.prices.opus.inPerMtok and .prices.sonnet.inPerMtok and .prices.haiku.inPerMtok' "$TABLE" >/dev/null 2>&1; then
-  record_pass "cost-table.json has opus/sonnet/haiku prices"
-else
-  record_fail "cost-table.json missing one of opus/sonnet/haiku"
-fi
-
-# --- 2: missing-data path returns exit 2 ---
-tmp=$(mktemp -d)
-pushd "$tmp" >/dev/null
-set +e
-bash "$RENDERER" GHOST-1 >/dev/null 2>&1
-rc=$?
-set -e
-if [ "$rc" -eq 2 ]; then
-  record_pass "renderer exits 2 when no tracker data"
-else
-  record_fail "renderer should exit 2 with no data (got $rc)"
-fi
-popd >/dev/null
-rm -rf "$tmp"
-
-# --- 3: tracker fixture renders markdown ---
-tmp=$(mktemp -d)
-pushd "$tmp" >/dev/null
-mkdir -p .worktrees/DEMO-1
-cat > .worktrees/DEMO-1/phase-tracker.json <<JSON
-{"taskId":"DEMO-1","phases":{
-  "1":{"name":"1. Analysis","model":"opus","tokens_in":12480,"tokens_out":3210},
-  "3":{"name":"3. Dev","model":"sonnet","tokens_in":42310,"tokens_out":9880}
-}}
-JSON
-out=$(bash "$RENDERER" DEMO-1)
-popd >/dev/null
-rm -rf "$tmp"
-
-if grep -q '^### Cost Summary$' <<< "$out" && \
-   grep -q '1. Analysis' <<< "$out" && \
-   grep -q '3. Dev' <<< "$out" && \
-   grep -q '\*\*Total\*\*' <<< "$out"; then
-  record_pass "tracker fixture renders header/rows/total"
-else
-  record_fail "tracker fixture output missing required rows"
-fi
-
-if grep -qE '\$[0-9]+\.[0-9]{2}' <<< "$out"; then
-  record_pass "USD formatted to 2 dp"
-else
-  record_fail "USD not formatted to 2 dp"
-fi
-
-# --- 4: OTel spans fallback ---
-tmp=$(mktemp -d)
-pushd "$tmp" >/dev/null
-mkdir -p .worktrees/OTEL-1
-cat > /tmp/cost-spans.jsonl <<JSON
-{"name":"phase.tokens.1","attributes":{"phase_id":"1","phase_name":"1. Analysis","model":"opus","tokens_in_delta":5000,"tokens_out_delta":1200}}
-{"name":"phase.tokens.1","attributes":{"phase_id":"1","phase_name":"1. Analysis","model":"opus","tokens_in_delta":3000,"tokens_out_delta":800}}
-JSON
-out2=$(bash "$RENDERER" OTEL-1 --otel-spans /tmp/cost-spans.jsonl || true)
-popd >/dev/null
-rm -rf "$tmp" /tmp/cost-spans.jsonl
-
-if grep -q '1. Analysis' <<< "$out2" && grep -q '8,000' <<< "$out2"; then
-  record_pass "OTel spans fallback aggregates phase tokens"
-else
-  record_fail "OTel fallback did not aggregate tokens (expected 8,000 in, got: $(echo "$out2" | tr '\n' ' ' | cut -c1-200))"
-fi
-
-# --- 5: unknown model →  -  + footnote ---
-tmp=$(mktemp -d)
-pushd "$tmp" >/dev/null
-mkdir -p .worktrees/UNK-1
-cat > .worktrees/UNK-1/phase-tracker.json <<JSON
-{"taskId":"UNK-1","phases":{
-  "1":{"name":"1. Analysis","model":"mystery-model","tokens_in":1000,"tokens_out":500}
-}}
-JSON
-out3=$(bash "$RENDERER" UNK-1)
-popd >/dev/null
-rm -rf "$tmp"
-
-if grep -q '|  - ' <<< "$out3" && grep -q 'USD unavailable' <<< "$out3"; then
-  record_pass "unknown model renders dash + footnote"
-else
-  record_fail "unknown model path missing dash or footnote"
-fi
-
-# --- 6: prefs schema exposes costSummary toggle ---
-if jq -e '.properties.global.properties.reportContent.properties.costSummary' "$SCHEMA" >/dev/null 2>&1; then
-  record_pass "prefs schema exposes reportContent.costSummary"
-else
-  record_fail "prefs schema missing reportContent.costSummary"
-fi
-
-# --- 7: channels.md documents the option ---
-if grep -q '^### Cost Summary' "$CHANNELS_DOC" && grep -q 'Cost özeti' "$CHANNELS_DOC"; then
-  record_pass "channels.md documents Cost Summary option"
-else
-  record_fail "channels.md missing Cost Summary docs"
-fi
-
-printf '\n══ cost-summary 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

package/pipeline/scripts/smoke-cross-phase-cohesion.sh

@@ -1,128 +0,0 @@
-#!/usr/bin/env bash
-#
-# smoke-cross-phase-cohesion.sh  -  assert every Phase N → Phase N+1 hand-off
-# in the multi-agent pipeline references the same schema on both sides.
-#
-# Why this exists (v8.0.0): existing schema-validation smoke proves each
-# JSON schema is itself well-formed. It does NOT prove that Phase 1's
-# reference doc says "I emit analysis-output.schema.json" while Phase 2's
-# reference doc says "I read analysis-output.schema.json". A reference can
-# rename or drop the schema without the validator catching it  -  phase docs
-# would silently desynchronize from the runtime contract.
-#
-# The contract chain:
-#   Phase 1 (Analysis)  →  analysis-output.schema.json    → Phase 2 (Planning)
-#   Phase 2 (Planning)  →  planning-output.schema.json    → Phase 3 (Dev)
-#   Phase 4 (Review)    →  reviewer-output.schema.json    → Phase 4 triage
-#   Phase 4 (Review)    →  triage-output.schema.json      → Phase 6 (Commit)
-#   Phase 6 (Commit)    →  agent-state.json fields        → Phase 7 (Report)
-#
-# Each schema must (a) exist on disk, (b) be referenced in the producer
-# phase ref doc, (c) be referenced in the consumer phase ref doc.
-#
-# Exit codes: 0=pass, 1=cohesion break detected
-set -euo pipefail
-
-REPO_ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-PHASE_DIR="$REPO_ROOT/pipeline/multi-agent-refs/phases"
-SCHEMA_DIR="$REPO_ROOT/pipeline/schemas"
-
-PASS=0
-FAIL=0
-pass() { echo "  ✓ $*"; PASS=$((PASS + 1)); }
-fail() { echo "  ✗ $*"; FAIL=$((FAIL + 1)); }
-
-assert_schema_exists() {
-  local schema="$1"
-  if [ -f "$SCHEMA_DIR/$schema" ]; then
-    pass "$schema exists in pipeline/schemas/"
-  else
-    fail "$schema MISSING  -  runtime contract broken"
-  fi
-}
-
-assert_phase_references() {
-  local phase_file="$1" schema="$2" role="$3"
-  local path="$PHASE_DIR/$phase_file"
-  if [ ! -f "$path" ]; then
-    fail "$phase_file not found at $path"
-    return
-  fi
-  if grep -Fq "$schema" "$path"; then
-    pass "$phase_file [$role] references $schema"
-  else
-    fail "$phase_file [$role] does NOT reference $schema  -  cohesion break"
-  fi
-}
-
-# -----------------------------------------------------------------------------
-echo "→ 1. Pipeline schemas all exist on disk"
-# -----------------------------------------------------------------------------
-assert_schema_exists "analysis-output.schema.json"
-assert_schema_exists "planning-output.schema.json"
-assert_schema_exists "reviewer-output.schema.json"
-assert_schema_exists "triage-output.schema.json"
-assert_schema_exists "agent-state.schema.json"
-assert_schema_exists "prefs.schema.json"
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "→ 2. Phase 1 → Phase 2 hand-off (analysis-output.schema.json)"
-# -----------------------------------------------------------------------------
-assert_phase_references "phase-1-analysis.md" "analysis-output.schema.json" "producer"
-assert_phase_references "phase-2-planning.md" "analysis-output.schema.json" "consumer"
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "→ 3. Phase 2 → Phase 3 hand-off (planning-output.schema.json)"
-# -----------------------------------------------------------------------------
-assert_phase_references "phase-2-planning.md" "planning-output.schema.json" "producer"
-assert_phase_references "phase-3-dev.md" "planning-output.schema.json" "consumer"
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "→ 4. Phase 4 review → triage hand-off (reviewer-output.schema.json)"
-# -----------------------------------------------------------------------------
-assert_phase_references "phase-4-review.md" "reviewer-output.schema.json" "producer"
-# Triage lives inside phase-4-review.md too; same file is producer + consumer for the chain.
-if grep -B 2 -A 60 "reviewer-output.schema.json" "$PHASE_DIR/phase-4-review.md" \
-   | grep -q "triage-output.schema.json"; then
-  pass "phase-4-review.md threads reviewer→triage in the same doc"
-else
-  fail "phase-4-review.md missing triage hand-off block"
-fi
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "→ 5. Phase 4 triage → Phase 6 commit hand-off (triage-output.schema.json)"
-# -----------------------------------------------------------------------------
-assert_phase_references "phase-4-review.md" "triage-output.schema.json" "producer"
-# Phase 6 doesn't always re-state the schema name; instead it references the
-# triage outcome bucket counts. Accept either form.
-if grep -Eq "triage-output\.schema\.json|triage outcome|accepted findings" "$PHASE_DIR/phase-6-commit.md"; then
-  pass "phase-6-commit.md consumes triage output (schema or outcome reference)"
-else
-  fail "phase-6-commit.md does NOT consume Phase 4 triage output"
-fi
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "→ 6. Phase 6 → Phase 7 hand-off (agent-state.json mutations)"
-# -----------------------------------------------------------------------------
-assert_phase_references "phase-6-commit.md" "agent-state" "producer"
-assert_phase_references "phase-7-report.md" "agent-state" "consumer"
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "→ 7. Validator runs the round-trip without errors"
-# -----------------------------------------------------------------------------
-if node "$REPO_ROOT/pipeline/scripts/validate-schemas.mjs" >/dev/null 2>&1; then
-  pass "validate-schemas.mjs reports all schemas valid"
-else
-  fail "validate-schemas.mjs failed  -  schema integrity broken"
-fi
-
-# -----------------------------------------------------------------------------
-echo ""
-echo "══ smoke-cross-phase-cohesion: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -gt 0 ] && exit 1 || exit 0

package/pipeline/scripts/smoke-description-tr.sh

@@ -1,82 +0,0 @@
-#!/usr/bin/env bash
-# smoke-description-tr.sh  -  localized picker-description contract.
-#
-# Verifies: every shipped command SKILL.md carries a description-tr line
-# (coverage), repo description: lines stay English with no description-en
-# sidecars committed (hygiene), and localize-commands.mjs apply/restore is
-# an idempotent, byte-identical round trip (mechanics).
-
-set -uo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
-REPO_ROOT="$(cd "$SCRIPT_DIR/../.." && pwd)"
-CMD_DIR="$REPO_ROOT/pipeline/commands/multi-agent"
-LOCALIZER="$REPO_ROOT/pipeline/scripts/localize-commands.mjs"
-
-PASS=0
-FAIL=0
-record_pass() { PASS=$((PASS + 1)); printf '  \342\234\223 %s\n' "$1"; }
-record_fail() { FAIL=$((FAIL + 1)); printf '  \342\234\227 %s\n' "$1"; }
-
-printf '\342\206\222 smoke-description-tr: localized picker-description contract\n'
-
-# 1. Coverage: every command SKILL.md (incl. the dispatcher) has description-tr
-missing=""
-count=0
-for f in "$CMD_DIR/SKILL.md" "$CMD_DIR"/*/SKILL.md; do
-  [ -f "$f" ] || continue
-  count=$((count + 1))
-  grep -q '^description-tr:' "$f" || missing="$missing ${f#"$CMD_DIR/"}"
-done
-if [ -z "$missing" ]; then
-  record_pass "all $count command SKILL.md files carry description-tr"
-else
-  record_fail "missing description-tr in:$missing"
-fi
-
-# 2. Hygiene: repo descriptions stay English, no sidecars committed
-if grep -rn '^description:.*[çğıöşüÇĞİÖŞÜ]' "$CMD_DIR" >/dev/null 2>&1; then
-  record_fail "repo description: lines must stay English (Turkish characters found)"
-else
-  record_pass "repo description: lines are English"
-fi
-if grep -rln '^description-en:' "$CMD_DIR" >/dev/null 2>&1; then
-  record_fail "description-en sidecar committed to the repo (installed-tree artifact)"
-else
-  record_pass "no description-en sidecar in the repo"
-fi
-
-# 3. Translation length within the agent-skills description budget
-long=$(awk '/^description-tr:/ && length($0) > 1044 { print FILENAME }' "$CMD_DIR"/*/SKILL.md "$CMD_DIR/SKILL.md" 2>/dev/null | head -1)
-if [ -z "$long" ]; then
-  record_pass "every description-tr is within the 1024-char budget"
-else
-  record_fail "description-tr exceeds budget in $long"
-fi
-
-# 4. Round trip: apply is idempotent, restore is byte-identical
-tmpdir=$(mktemp -d)
-trap 'rm -rf "$tmpdir"' EXIT
-cp -R "$CMD_DIR" "$tmpdir/cmds"
-node "$LOCALIZER" apply --dir "$tmpdir/cmds" >/dev/null 2>&1
-if grep -q '^description:.*[çğıöşüÇĞİÖŞÜ]' "$tmpdir/cmds/dev/SKILL.md" \
-  && grep -q '^description-en:' "$tmpdir/cmds/dev/SKILL.md"; then
-  record_pass "apply swaps description and writes the description-en sidecar"
-else
-  record_fail "apply did not localize the sample file"
-fi
-second=$(node "$LOCALIZER" apply --dir "$tmpdir/cmds" 2>/dev/null)
-if printf '%s' "$second" | grep -q ' 0 file(s) changed'; then
-  record_pass "second apply is a no-op (idempotent)"
-else
-  record_fail "second apply changed files again: $second"
-fi
-node "$LOCALIZER" restore --dir "$tmpdir/cmds" >/dev/null 2>&1
-if diff -r "$CMD_DIR" "$tmpdir/cmds" >/dev/null 2>&1; then
-  record_pass "restore returns a byte-identical tree"
-else
-  record_fail "restore left differences vs the original tree"
-fi
-
-printf '\n\342\225\220\342\225\220 smoke-description-tr: %d passed, %d failed \342\225\220\342\225\220\n' "$PASS" "$FAIL"
-[ "$FAIL" -eq 0 ]

package/pipeline/scripts/smoke-dev-critic.sh

@@ -1,144 +0,0 @@
-#!/usr/bin/env bash
-# smoke-dev-critic.sh
-#
-# Verifies the Phase 3.5 evaluator-optimizer contract:
-#   1. pipeline/agents/dev-critic.md exists with required frontmatter
-#   2. phase-3-dev.md documents Step 3.5 with loop cap + gates + escalate rule
-#   3. prefs.schema.json exposes global.devCritic.{enabled,maxIterations,model,minDiffLoc}
-#   4. dev-critic-output.schema.json exists and is valid JSON Schema
-#   5. Severity enum + gate enum match the agent doc
-#   6. Loop cap default is 2 (matches "max 2" rule documented in the agent)
-#   7. Off-by-default (no surprise cost on existing users)
-#
-# Exit 0 = all pass, 1 = any failure.
-
-set -euo pipefail
-
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-AGENT="$ROOT/pipeline/agents/dev-critic.md"
-PHASE_DOC="$ROOT/pipeline/multi-agent-refs/phases/phase-3-dev.md"
-PREFS_SCHEMA="$ROOT/pipeline/schemas/prefs.schema.json"
-CRITIC_SCHEMA="$ROOT/pipeline/schemas/dev-critic-output.schema.json"
-
-pass=0
-fail=0
-failures=()
-record_pass() { pass=$((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' "$1"; }
-
-printf '→ smoke-dev-critic: Phase 3.5 evaluator-optimizer contract\n'
-
-# 1. Agent file exists with required frontmatter
-if [ ! -f "$AGENT" ]; then
-  record_fail "pipeline/agents/dev-critic.md missing"
-else
-  # Extract frontmatter (between first two --- lines)
-  fm=$(awk '/^---$/{n++; if(n==2) exit; if(n==1) next} n==1{print}' "$AGENT")
-  for key in description model preferredModel; do
-    if grep -qE "^${key}:" <<<"$fm"; then
-      record_pass "agent frontmatter has '${key}'"
-    else
-      record_fail "agent frontmatter missing '${key}'"
-    fi
-  done
-  # Model must be sonnet (the documented default)
-  if grep -qE "^model: sonnet" <<<"$fm"; then
-    record_pass "agent default model is sonnet"
-  else
-    record_fail "agent model default should be sonnet (was: $(grep '^model:' <<<"$fm" || echo missing))"
-  fi
-fi
-
-# 2. Phase doc mentions Step 3.5 with the loop contract
-if [ ! -f "$PHASE_DOC" ]; then
-  record_fail "phase-3-dev.md missing"
-else
-  if grep -qF "Step 3.5" "$PHASE_DOC" && grep -qF "Dev Critic" "$PHASE_DOC"; then
-    record_pass "phase-3-dev.md documents Step 3.5  -  Dev Critic"
-  else
-    record_fail "phase-3-dev.md missing Step 3.5 section"
-  fi
-  if grep -qE "max(imum)? (\*\*)?2(\*\*)? iterations?" "$PHASE_DOC"; then
-    record_pass "phase doc documents max 2 iterations"
-  else
-    record_fail "phase doc must state 'max 2 iterations' rule"
-  fi
-  if grep -qF "escalate" "$PHASE_DOC"; then
-    record_pass "phase doc documents escalate path"
-  else
-    record_fail "phase doc must document escalate behavior"
-  fi
-  if grep -qF "dev-critic.md" "$PHASE_DOC"; then
-    record_pass "phase doc cites agents/dev-critic.md"
-  else
-    record_fail "phase doc must reference agents/dev-critic.md"
-  fi
-fi
-
-# 3. Prefs schema exposes devCritic toggle
-if jq -e '.properties.global.properties.devCritic.properties.enabled' "$PREFS_SCHEMA" >/dev/null 2>&1; then
-  record_pass "prefs schema exposes devCritic.enabled"
-else
-  record_fail "prefs schema missing devCritic.enabled"
-fi
-for prop in maxIterations model minDiffLoc; do
-  if jq -e ".properties.global.properties.devCritic.properties.${prop}" "$PREFS_SCHEMA" >/dev/null 2>&1; then
-    record_pass "prefs schema exposes devCritic.${prop}"
-  else
-    record_fail "prefs schema missing devCritic.${prop}"
-  fi
-done
-
-# 4. dev-critic-output schema is valid JSON
-if [ ! -f "$CRITIC_SCHEMA" ]; then
-  record_fail "dev-critic-output.schema.json missing"
-elif ! jq empty "$CRITIC_SCHEMA" 2>/dev/null; then
-  record_fail "dev-critic-output.schema.json is not valid JSON"
-else
-  record_pass "dev-critic-output.schema.json parses"
-fi
-
-# 5. Severity enum matches doc (blocking/important/suggestion)
-if jq -e '.properties.findings.items.properties.severity.enum
-          | (index("blocking") != null
-             and index("important") != null
-             and index("suggestion") != null)' "$CRITIC_SCHEMA" >/dev/null 2>&1; then
-  record_pass "schema severity enum includes blocking/important/suggestion"
-else
-  record_fail "schema severity enum incomplete"
-fi
-
-# Gate keys (build/lint/test/secrets) match doc
-for gate in build lint test secrets; do
-  if jq -e ".properties.gates.properties.${gate}" "$CRITIC_SCHEMA" >/dev/null 2>&1; then
-    record_pass "schema gates expose '${gate}'"
-  else
-    record_fail "schema gates missing '${gate}'"
-  fi
-done
-
-# 6. Loop cap default = 2
-default_iter=$(jq -r '.properties.global.properties.devCritic.properties.maxIterations.default // empty' "$PREFS_SCHEMA")
-if [ "$default_iter" = "2" ]; then
-  record_pass "devCritic.maxIterations defaults to 2"
-else
-  record_fail "devCritic.maxIterations default should be 2 (was: ${default_iter:-missing})"
-fi
-
-# 7. Off by default  -  preserves existing-user baseline
-# Note: jq's `// empty` treats boolean false as falsy, so use `has("default")` + raw value.
-if jq -e '.properties.global.properties.devCritic.properties.enabled
-          | has("default") and .default == false' "$PREFS_SCHEMA" >/dev/null 2>&1; then
-  record_pass "devCritic.enabled defaults to false (opt-in)"
-else
-  default_enabled=$(jq -c '.properties.global.properties.devCritic.properties.enabled.default' "$PREFS_SCHEMA")
-  record_fail "devCritic.enabled should default to false (was: ${default_enabled})"
-fi
-
-printf '\n══ dev-critic smoke: %d passed, %d failed ══\n' "$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

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

@@ -1,147 +0,0 @@
-#!/usr/bin/env bash
-# smoke-diff-explain.sh  -  v7.8.0 Paket B: Phase 4 triage to diff bridge.
-#
-# Validates:
-#   1. Slash command + Copilot peer present (parity smoke also catches this)
-#   2. Empty triage produces a "no findings" stub without crashing
-#   3. Mixed-classification fixture renders all bucket icons (✅ ⏭️ ❌)
-#   4. Severity icons (🚫 ⚠️ 💡) render correctly
-#   5. Hunk binding: synthetic diff + matching line → diff fence in output
-#   6. Out-of-diff line → fallback message (no crash)
-#   7. --help works without inputs
-#   8. Missing inputs exit 1 with a clear error
-#   9. Read-only contract: script does NOT call git apply / git checkout / write APIs
-
-set -uo pipefail
-
-REPO_ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
-SCRIPT="$REPO_ROOT/pipeline/scripts/diff-explain.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)
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 1. Command + skill files exist"
-[ -f "$REPO_ROOT/pipeline/commands/multi-agent/diff-explain/SKILL.md" ] && pass "diff-explain.md present" || fail "diff-explain.md missing"
-[ -f "$REPO_ROOT/pipeline/skills/shared/core/multi-agent-diff-explain/SKILL.md" ] && pass "multi-agent-diff-explain SKILL.md present" || fail "Copilot peer missing"
-[ -x "$SCRIPT" ] && pass "diff-explain.mjs is executable" || fail "diff-explain.mjs not executable"
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 2. Empty findings stub"
-cat > "$TMP/empty.json" <<'EOF'
-{"accepted": [], "deferred": [], "rejected": [], "approved": true}
-EOF
-OUT=$(node "$SCRIPT" --triage "$TMP/empty.json" --diff /dev/null 2>&1)
-if echo "$OUT" | grep -q "no findings to map"; then
-  pass "empty triage → 'no findings' stub"
-else
-  fail "empty triage did not produce stub message"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 3. Bucket icons render correctly"
-OUT=$(node "$SCRIPT" --triage "$REPO_ROOT/pipeline/eval/triage/05-mixed-classification/expected.json" --diff /dev/null 2>&1)
-echo "$OUT" | grep -q "✅" && pass "accepted bucket icon (✅) rendered" || fail "accepted icon missing"
-echo "$OUT" | grep -q "⏭️" && pass "deferred bucket icon (⏭️) rendered" || fail "deferred icon missing"
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 4. Severity icons render"
-echo "$OUT" | grep -q "🚫" && pass "blocking severity icon (🚫) rendered" || fail "blocking icon missing"
-echo "$OUT" | grep -q "⚠️" && pass "important severity icon (⚠️) rendered" || fail "important icon missing"
-echo "$OUT" | grep -q "💡" && pass "suggestion severity icon (💡) rendered" || fail "suggestion icon missing"
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 5. Hunk binding  -  synthetic diff matches finding line"
-cat > "$TMP/syn.diff" <<'EOF'
-diff --git a/src/api/checkout.ts b/src/api/checkout.ts
-index a..b 100644
---- a/src/api/checkout.ts
-+++ b/src/api/checkout.ts
-@@ -98,5 +98,7 @@ ...
-   const total = sum(items);
-+  const total2 = sum(items);
-+  // duplicate
-   await chargeCard(total);
-EOF
-OUT=$(node "$SCRIPT" --triage "$REPO_ROOT/pipeline/eval/triage/05-mixed-classification/expected.json" --diff "$TMP/syn.diff" 2>&1)
-if echo "$OUT" | grep -q '```diff'; then
-  pass "hunk fenced diff block rendered"
-else
-  fail "no diff code fence in output"
-fi
-if echo "$OUT" | grep -q "@@ -98,5 +98,7 @@"; then
-  pass "hunk header preserved in output"
-else
-  fail "hunk header missing in output"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 6. Out-of-diff line fallback"
-# checkout.ts is in the diff but its lines 145, 200 are NOT covered by the
-# single hunk above. Should produce fallback messages, not crash.
-fallback_count=$(echo "$OUT" | grep -c "line not in diff" || true)
-if [ "$fallback_count" -ge 1 ]; then
-  pass "out-of-diff lines produce fallback message ($fallback_count occurrences)"
-else
-  # Acceptable alternative: hunk renderer outputs nothing visible. Both options OK.
-  pass "out-of-diff lines handled (no crash)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 7. --help works without inputs"
-HELP=$(node "$SCRIPT" --help 2>&1)
-if echo "$HELP" | grep -q "Usage:"; then
-  pass "--help prints usage"
-else
-  fail "--help missing Usage block"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 8. Missing inputs exit non-zero"
-set +e
-node "$SCRIPT" 2>"$TMP/err.txt"
-EC=$?
-set -e
-if [ "$EC" -ne 0 ] && grep -q "required" "$TMP/err.txt"; then
-  pass "no-input exit $EC with 'required' error"
-else
-  fail "no-input did not error correctly (exit=$EC)"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 9. Read-only contract  -  no destructive git APIs"
-if grep -qE "git apply|git checkout|git reset|git rebase|git push|git commit|writeFileSync" "$SCRIPT"; then
-  fail "diff-explain.mjs references destructive git API or writeFileSync (read-only contract violated)"
-else
-  pass "no destructive git/write API references in diff-explain.mjs"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo "→ 10. Injection-safe  -  shell metacharacters in --base are not executed"
-SENTINEL="$TMP/PWNED"
-set +e
-node "$SCRIPT" \
-  --triage "$REPO_ROOT/pipeline/eval/triage/05-mixed-classification/expected.json" \
-  --base "main; touch $SENTINEL" --branch HEAD >/dev/null 2>&1
-set -e
-if [ -f "$SENTINEL" ]; then
-  fail "command injection: --base executed arbitrary shell (sentinel created)"
-else
-  pass "shell metacharacters in --base not executed (execFileSync)"
-fi
-if grep -q "execSync" "$SCRIPT"; then
-  fail "diff-explain.mjs still uses execSync (shell interpolation risk)"
-else
-  pass "diff-explain.mjs uses execFileSync only"
-fi
-
-# ──────────────────────────────────────────────────────────────────────────
-echo ""
-echo "══ diff-explain smoke: $PASS passed, $FAIL failed ══"
-[ "$FAIL" -eq 0 ]