cc-safe-setup 29.8.0 → 30.0.5

package/tests/test-nested-spawn-inflight-guard.sh

@@ -0,0 +1,281 @@
+#!/bin/bash
+# Tests for nested-spawn-inflight-guard.sh
+HOOK="$(dirname "$0")/../examples/nested-spawn-inflight-guard.sh"
+PASS=0
+FAIL=0
+
+mktempd() {
+  mktemp -d "${TMPDIR:-/tmp}/cc-nested-spawn-test.XXXXXX"
+}
+
+run_test() {
+  local desc="$1"
+  local result="$2"
+  if [ "$result" = "pass" ]; then
+    echo "  PASS: $desc"
+    PASS=$((PASS + 1))
+  else
+    echo "  FAIL: $desc"
+    FAIL=$((FAIL + 1))
+  fi
+}
+
+# Build a transcript file with N in-flight Task dispatches.
+# Args: $1=output_path  $2=count_of_dispatched  $3=count_of_resolved
+mk_transcript() {
+  local path="$1"
+  local dispatched="$2"
+  local resolved="$3"
+  : > "$path"
+  local i=0
+  while [ "$i" -lt "$dispatched" ]; do
+    printf '{"message":{"content":[{"type":"tool_use","id":"toolu_%03d","name":"Task","input":{"description":"d"}}]}}\n' "$i" >> "$path"
+    i=$((i + 1))
+  done
+  i=0
+  while [ "$i" -lt "$resolved" ]; do
+    printf '{"message":{"content":[{"type":"tool_result","tool_use_id":"toolu_%03d","content":"done"}]}}\n' "$i" >> "$path"
+    i=$((i + 1))
+  done
+}
+
+echo "Testing nested-spawn-inflight-guard.sh"
+echo "======================================"
+
+# Test 1: non-dispatcher tool (Bash) → silent exit 0
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 10 0
+EXIT=0
+echo "{\"tool_name\":\"Bash\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "non-dispatcher tool name → exit 0" pass || run_test "non-dispatcher tool name → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 2: non-dispatcher tool (Read) → silent exit 0
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 10 0
+echo "{\"tool_name\":\"Read\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "Read tool → exit 0" pass || run_test "Read tool → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 3: missing transcript_path → fail open exit 0
+echo '{"tool_name":"Task"}' | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "missing transcript_path → fail open exit 0" pass || run_test "missing transcript_path → fail open exit 0 (got $EXIT)" fail
+
+# Test 4: unreadable transcript_path → fail open exit 0
+echo '{"tool_name":"Task","transcript_path":"/nonexistent/path.jsonl"}' | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "unreadable transcript → fail open exit 0" pass || run_test "unreadable transcript → fail open exit 0 (got $EXIT)" fail
+
+# Test 5: CC_NESTED_SPAWN_DISABLE=1 → bypass even when budget exceeded
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 20 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | CC_NESTED_SPAWN_DISABLE=1 bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "CC_NESTED_SPAWN_DISABLE=1 bypasses budget" pass || run_test "CC_NESTED_SPAWN_DISABLE=1 bypasses budget (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 6: CC_NESTED_SPAWN_OVERRIDE=1 → bypass even when budget exceeded
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 20 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | CC_NESTED_SPAWN_OVERRIDE=1 bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "CC_NESTED_SPAWN_OVERRIDE=1 bypasses budget" pass || run_test "CC_NESTED_SPAWN_OVERRIDE=1 bypasses budget (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 7: zero in-flight (5 dispatched, 5 resolved) → exit 0
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 5 5
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "zero in-flight (5d/5r) → exit 0" pass || run_test "zero in-flight (5d/5r) → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 8: 3 in-flight (5 dispatched, 2 resolved), budget 5 → exit 0
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 5 2
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "3 in-flight below budget 5 → exit 0" pass || run_test "3 in-flight below budget 5 → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 9: 5 in-flight (5 dispatched, 0 resolved), budget 5 → exit 2 (at budget)
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 5 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "5 in-flight at budget 5 → exit 2" pass || run_test "5 in-flight at budget 5 → exit 2 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 10: 7 in-flight, budget 5 → exit 2 (above budget)
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 7 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "7 in-flight above budget 5 → exit 2" pass || run_test "7 in-flight above budget 5 → exit 2 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 11: custom budget via env var (CC_NESTED_SPAWN_BUDGET=3, 4 in-flight) → exit 2
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 4 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | CC_NESTED_SPAWN_BUDGET=3 bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "custom budget 3, 4 in-flight → exit 2" pass || run_test "custom budget 3, 4 in-flight → exit 2 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 12: custom budget allows more (budget=10, 7 in-flight) → exit 0
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 7 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | CC_NESTED_SPAWN_BUDGET=10 bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "custom budget 10, 7 in-flight → exit 0" pass || run_test "custom budget 10, 7 in-flight → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 13: legacy "Agent" tool name matched
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4; do
+  printf '{"message":{"content":[{"type":"tool_use","id":"toolu_a%d","name":"Agent","input":{}}]}}\n' "$i" >> "$TX"
+done
+echo "{\"tool_name\":\"Agent\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "legacy Agent tool name → tracked" pass || run_test "legacy Agent tool name → tracked (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 14: SendMessage tool name matched
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4; do
+  printf '{"message":{"content":[{"type":"tool_use","id":"toolu_s%d","name":"SendMessage","input":{}}]}}\n' "$i" >> "$TX"
+done
+echo "{\"tool_name\":\"SendMessage\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "SendMessage tool name → tracked" pass || run_test "SendMessage tool name → tracked (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 15: mixed tool calls (Task + Bash) — only Task counted
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4 5 6 7 8 9; do
+  # 10 Bash calls — should not count
+  printf '{"message":{"content":[{"type":"tool_use","id":"toolu_b%d","name":"Bash","input":{}}]}}\n' "$i" >> "$TX"
+done
+for i in 0 1; do
+  # 2 Task calls — under budget
+  printf '{"message":{"content":[{"type":"tool_use","id":"toolu_t%d","name":"Task","input":{}}]}}\n' "$i" >> "$TX"
+done
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "mixed tool calls — only Task family counted" pass || run_test "mixed tool calls — only Task family counted (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 16: malformed JSON in transcript → fail open (exit 0)
+TD=$(mktempd); TX="$TD/t.jsonl"
+echo "not valid json" > "$TX"
+echo "this is also broken {" >> "$TX"
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "malformed transcript → fail open" pass || run_test "malformed transcript → fail open (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 17: advisory message names budget and in-flight ids
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 6 0
+STDERR=$(echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" 2>&1 >/dev/null)
+EXIT_NAMED=0
+echo "$STDERR" | grep -q "BLOCKED: 6 subagent" || EXIT_NAMED=1
+echo "$STDERR" | grep -q "budget: 5" || EXIT_NAMED=1
+echo "$STDERR" | grep -q "toolu_000" || EXIT_NAMED=1
+echo "$STDERR" | grep -q "CC_NESTED_SPAWN_OVERRIDE" || EXIT_NAMED=1
+[ "$EXIT_NAMED" = "0" ] && run_test "advisory names count, budget, ids, override path" pass || run_test "advisory names count, budget, ids, override path (missing fields)" fail
+rm -rf "$TD"
+
+# Test 18: non-numeric budget → fail open
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 10 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | CC_NESTED_SPAWN_BUDGET=abc bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "non-numeric budget → fail open" pass || run_test "non-numeric budget → fail open (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 19: tool_use_id reuse — last write wins (in-flight only if no later result)
+# 3 dispatched (id 0..2), then 2 resolved (id 0,1), id 2 still in-flight → 1 in-flight, budget 5 → pass
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 3 2
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "3 dispatched / 2 resolved → 1 in-flight → exit 0" pass || run_test "3 dispatched / 2 resolved → 1 in-flight → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 20: LOOKBACK truncation does not over-count old completed work
+# Old work has 100 dispatched + 100 resolved early in the file; only last 10 lines visible
+# In the visible window: 5 dispatched, 0 resolved → 5 in-flight, budget 5 → exit 2
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4 5 6 7 8 9; do
+  printf '{"message":{"content":[{"type":"tool_use","id":"old_%d","name":"Task","input":{}}]}}\n' "$i" >> "$TX"
+  printf '{"message":{"content":[{"type":"tool_result","tool_use_id":"old_%d","content":"d"}]}}\n' "$i" >> "$TX"
+done
+for i in 100 101 102 103 104; do
+  printf '{"message":{"content":[{"type":"tool_use","id":"recent_%d","name":"Task","input":{}}]}}\n' "$i" >> "$TX"
+done
+# LOOKBACK=5 makes only the last 5 lines visible — all 5 are recent_* tool_use entries.
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | CC_NESTED_SPAWN_LOOKBACK=5 bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "LOOKBACK=5 sees only recent 5 in-flight → exit 2" pass || run_test "LOOKBACK=5 sees only recent 5 in-flight → exit 2 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 21: tool family in MCP-style entries (alternate transcript shape: bare tool_use object, no message wrapper)
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4; do
+  printf '{"type":"tool_use","id":"bare_%d","name":"Task","input":{}}\n' "$i" >> "$TX"
+done
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "bare tool_use shape (no message wrapper) → tracked" pass || run_test "bare tool_use shape (no message wrapper) → tracked (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 22: tool_name with unrelated name like "TaskComplete" must NOT match "Task" prefix
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4 5 6 7 8 9; do
+  printf '{"message":{"content":[{"type":"tool_use","id":"tc_%d","name":"TaskComplete","input":{}}]}}\n' "$i" >> "$TX"
+done
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "TaskComplete (different tool) not counted as Task" pass || run_test "TaskComplete (different tool) not counted as Task (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 23: tool_name with case mismatch ("task" lowercase) must NOT match (exact case)
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+for i in 0 1 2 3 4; do
+  printf '{"message":{"content":[{"type":"tool_use","id":"lc_%d","name":"task","input":{}}]}}\n' "$i" >> "$TX"
+done
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "lowercase 'task' (case mismatch) not counted" pass || run_test "lowercase 'task' (case mismatch) not counted (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 24: assistant turn with multiple tool_use in one content array (parallel dispatch)
+TD=$(mktempd); TX="$TD/t.jsonl"
+: > "$TX"
+# One assistant turn that dispatches 6 Task calls in parallel
+printf '{"message":{"content":[' >> "$TX"
+for i in 0 1 2 3 4 5; do
+  [ "$i" -gt 0 ] && printf ',' >> "$TX"
+  printf '{"type":"tool_use","id":"par_%d","name":"Task","input":{}}' "$i" >> "$TX"
+done
+printf ']}}\n' >> "$TX"
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "2" ] && run_test "single turn with 6 parallel Task dispatches → exit 2" pass || run_test "single turn with 6 parallel Task dispatches → exit 2 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 25: empty transcript file (zero lines) → no in-flight → exit 0
+TD=$(mktempd); TX="$TD/t.jsonl"; : > "$TX"
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "empty transcript → 0 in-flight → exit 0" pass || run_test "empty transcript → 0 in-flight → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+# Test 26: 4 in-flight default budget 5 → still safe (exit 0)
+TD=$(mktempd); TX="$TD/t.jsonl"; mk_transcript "$TX" 4 0
+echo "{\"tool_name\":\"Task\",\"transcript_path\":\"$TX\"}" | bash "$HOOK" >/dev/null 2>&1
+EXIT=$?
+[ "$EXIT" = "0" ] && run_test "4 in-flight under default budget 5 → exit 0" pass || run_test "4 in-flight under default budget 5 → exit 0 (got $EXIT)" fail
+rm -rf "$TD"
+
+echo "========================================"
+echo "Results: $PASS passed, $FAIL failed"
+[ "$FAIL" = "0" ] && exit 0 || exit 1

package/tests/test-no-force-flag.sh

@@ -0,0 +1,49 @@
+#!/bin/bash
+# Tests for no-force-flag.sh
+# Run: bash tests/test-no-force-flag.sh
+set -uo pipefail
+
+HOOK="$(dirname "$0")/../examples/no-force-flag.sh"
+PASS=0 FAIL=0
+
+run_test() {
+    local desc="$1" expected_exit="$2" cmd="$3"
+    local actual_exit
+    local input="{\"tool_name\":\"Bash\",\"tool_input\":{\"command\":${cmd}}}"
+    echo "$input" | bash "$HOOK" >/dev/null 2>/dev/null
+    actual_exit=$?
+    if [ "$actual_exit" -eq "$expected_exit" ]; then
+        echo "  PASS: $desc"
+        PASS=$((PASS + 1))
+    else
+        echo "  FAIL: $desc (expected exit $expected_exit, got $actual_exit)"
+        FAIL=$((FAIL + 1))
+    fi
+}
+
+echo "Testing no-force-flag.sh"
+echo "========================"
+
+# --- BLOCK: git push --force (long form) ---
+run_test "git push --force blocks"            2 '"git push --force"'
+run_test "git push origin main --force blocks" 2 '"git push origin main --force"'
+
+# --- BLOCK: git push -f (short form — previously bypassed entirely) ---
+run_test "git push -f blocks"                 2 '"git push -f"'
+run_test "git push origin main -f blocks"     2 '"git push origin main -f"'
+run_test "git push -f origin feature blocks"  2 '"git push -f origin feature"'
+
+# --- ALLOW: --force-with-lease and non-force pushes (no false positives) ---
+run_test "git push --force-with-lease allows" 0 '"git push --force-with-lease"'
+run_test "git push allows"                    0 '"git push"'
+run_test "git push origin main allows"        0 '"git push origin main"'
+run_test "git push to branch ending in -f allows" 0 '"git push origin feature-f"'
+
+# --- BLOCK: npm install --force / docker prune -f (other patterns, unchanged) ---
+run_test "npm install --force blocks"         2 '"npm install --force"'
+run_test "docker system prune -f blocks"      2 '"docker system prune -f"'
+
+echo
+echo "========================"
+echo "Results: $PASS passed, $FAIL failed"
+exit $FAIL

package/tests/test-non-english-quality-warner.sh

@@ -0,0 +1,229 @@
+#!/bin/bash
+# Tests for non-english-quality-warner.sh
+HOOK="$(dirname "$0")/../examples/non-english-quality-warner.sh"
+PASS=0
+FAIL=0
+
+run_test() {
+  local desc="$1"
+  local result="$2"
+  if [ "$result" = "pass" ]; then
+    echo "  PASS: $desc"
+    PASS=$((PASS + 1))
+  else
+    echo "  FAIL: $desc"
+    FAIL=$((FAIL + 1))
+  fi
+}
+
+reset_env() {
+  unset CC_NON_ENGLISH_QUALITY_WARNER_REMIND
+  unset CC_NON_ENGLISH_QUALITY_WARNER_DISABLE
+  unset CC_NON_ENGLISH_QUALITY_WARNER_QUIET
+}
+
+echo "Testing non-english-quality-warner.sh"
+echo "======================================"
+
+# Test 1: Default (no env vars) → silent
+reset_env
+OUT=$(bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && [ -z "$OUT" ]; then
+  run_test "Default (no env vars) is silent" pass
+else
+  run_test "Default (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 2: REMIND=1 → emits advisory
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && printf '%s' "$OUT" | grep -q "non-english-quality-warner"; then
+  run_test "REMIND=1 emits the advisory" pass
+else
+  run_test "REMIND=1 (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 3: REMIND=1 + DISABLE=1 → silent (DISABLE wins)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 CC_NON_ENGLISH_QUALITY_WARNER_DISABLE=1 bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && [ -z "$OUT" ]; then
+  run_test "REMIND=1 + DISABLE=1 silent (DISABLE wins)" pass
+else
+  run_test "REMIND+DISABLE (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 4: REMIND=1 + QUIET=1 → silent (QUIET wins)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 CC_NON_ENGLISH_QUALITY_WARNER_QUIET=1 bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && [ -z "$OUT" ]; then
+  run_test "REMIND=1 + QUIET=1 silent (QUIET wins)" pass
+else
+  run_test "REMIND+QUIET (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 5: REMIND empty string → silent (not "1")
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND="" bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && [ -z "$OUT" ]; then
+  run_test "REMIND='' is silent (not '1')" pass
+else
+  run_test "REMIND='' (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 6: REMIND=0 → silent
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=0 bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && [ -z "$OUT" ]; then
+  run_test "REMIND=0 is silent" pass
+else
+  run_test "REMIND=0 (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 7: REMIND=true → silent (only literal "1" enables)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=true bash "$HOOK" 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && [ -z "$OUT" ]; then
+  run_test "REMIND=true is silent (only literal '1' enables)" pass
+else
+  run_test "REMIND=true (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 8: Advisory cites issue #62961 (Korean evidence)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if printf '%s' "$OUT" | grep -q "62961"; then
+  run_test "Advisory cites issue #62961" pass
+else
+  run_test "Advisory cites #62961 (out=$OUT)" fail
+fi
+
+# Test 9: Advisory cites 18× signal
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if printf '%s' "$OUT" | grep -q "18.0×"; then
+  run_test "Advisory cites 18× regression signal" pass
+else
+  run_test "Advisory cites 18× (out_len=${#OUT})" fail
+fi
+
+# Test 10: Advisory cites the three workaround paths
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+HAS1=$(printf '%s' "$OUT" | grep -c "1) Pin formal register")
+HAS2=$(printf '%s' "$OUT" | grep -c "2) Switch to Sonnet")
+HAS3=$(printf '%s' "$OUT" | grep -c "3) Post-process the output")
+if [ "$HAS1" = "1" ] && [ "$HAS2" = "1" ] && [ "$HAS3" = "1" ]; then
+  run_test "Advisory contains all three workaround paths" pass
+else
+  run_test "Three paths (has1=$HAS1 has2=$HAS2 has3=$HAS3)" fail
+fi
+
+# Test 11: Advisory references the version timeline (2.1.126, 2.1.132)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+H126=$(printf '%s' "$OUT" | grep -c "2.1.126")
+H132=$(printf '%s' "$OUT" | grep -c "2.1.132")
+if [ "$H126" -ge "1" ] && [ "$H132" -ge "1" ]; then
+  run_test "Advisory references version timeline 2.1.126 and 2.1.132" pass
+else
+  run_test "Version timeline (h126=$H126 h132=$H132)" fail
+fi
+
+# Test 12: Advisory shows how to disable (unset REMIND and/or QUIET=1)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if printf '%s' "$OUT" | grep -q "unset CC_NON_ENGLISH_QUALITY_WARNER_REMIND" \
+   && printf '%s' "$OUT" | grep -q "export CC_NON_ENGLISH_QUALITY_WARNER_QUIET=1"; then
+  run_test "Advisory explains how to disable" pass
+else
+  run_test "Disable instructions (out_len=${#OUT})" fail
+fi
+
+# Test 13: Advisory references the field guide gist
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if printf '%s' "$OUT" | grep -q "gist.github.com"; then
+  run_test "Advisory references the field guide gist" pass
+else
+  run_test "Field guide gist link (out_len=${#OUT})" fail
+fi
+
+# Test 14: Advisory includes Korean and Japanese formal-register examples
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+HAS_KO=$(printf '%s' "$OUT" | grep -c "박다")
+HAS_JA=$(printf '%s' "$OUT" | grep -c "formal register に固定")
+if [ "$HAS_KO" -ge "1" ] && [ "$HAS_JA" -ge "1" ]; then
+  run_test "Advisory contains Korean and Japanese examples" pass
+else
+  run_test "Multilingual examples (has_ko=$HAS_KO has_ja=$HAS_JA)" fail
+fi
+
+# Test 15: Advisory recommends Sonnet swap with concrete export command
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if printf '%s' "$OUT" | grep -q "export ANTHROPIC_MODEL=claude-sonnet-4-7"; then
+  run_test "Advisory shows concrete Sonnet swap command" pass
+else
+  run_test "Sonnet swap command (out_len=${#OUT})" fail
+fi
+
+# Test 16: Never blocks — exit 0 even when emitting
+reset_env
+CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>/dev/null
+EXIT=$?
+if [ "$EXIT" = "0" ]; then
+  run_test "Never blocks (exit=0 even when emitting)" pass
+else
+  run_test "Exit code on emit (exit=$EXIT)" fail
+fi
+
+# Test 17: Never blocks — exit 0 when silent (default)
+reset_env
+bash "$HOOK" 2>/dev/null
+EXIT=$?
+if [ "$EXIT" = "0" ]; then
+  run_test "Never blocks (exit=0 when silent)" pass
+else
+  run_test "Exit code when silent (exit=$EXIT)" fail
+fi
+
+# Test 18: Hook handles no stdin input gracefully (SessionStart hooks may not pipe)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" < /dev/null 2>&1)
+EXIT=$?
+if [ "$EXIT" = "0" ] && printf '%s' "$OUT" | grep -q "non-english-quality-warner"; then
+  run_test "Handles no-stdin gracefully (SessionStart pattern)" pass
+else
+  run_test "No-stdin (exit=$EXIT, out_len=${#OUT})" fail
+fi
+
+# Test 19: Advisory is non-trivial size (sanity check)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if [ "${#OUT}" -ge "1500" ]; then
+  run_test "Advisory has substantial content (>=1500 chars)" pass
+else
+  run_test "Advisory size (chars=${#OUT})" fail
+fi
+
+# Test 20: Advisory acknowledges the regression is server-side (no client fix)
+reset_env
+OUT=$(CC_NON_ENGLISH_QUALITY_WARNER_REMIND=1 bash "$HOOK" 2>&1)
+if printf '%s' "$OUT" | grep -q "server-side"; then
+  run_test "Advisory acknowledges server-side cause (no client fix)" pass
+else
+  run_test "Server-side framing (out_len=${#OUT})" fail
+fi
+
+echo "======================================"
+echo "PASS: $PASS"
+echo "FAIL: $FAIL"
+[ "$FAIL" = "0" ] && exit 0 || exit 1