@michelj/context-guard 0.4.2 → 0.4.3

package/tests/BC-20260709-006.sh

@@ -0,0 +1,76 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT="${CONTEXT_GUARD_SCRIPT:-$(cd "$(dirname "$0")/.." && pwd)/scripts/context_guard.py}"
+ROOT="$(mktemp -d "${TMPDIR:-/tmp}/context-guard-loose-bc-XXXXXX")"
+trap 'rm -rf "$ROOT"' EXIT
+
+mkdir -p "$ROOT/bin"
+cat >"$ROOT/bin/smoke.sh" <<'SH'
+#!/usr/bin/env bash
+set -euo pipefail
+test -f index.html
+grep -Fq "Evening Reset" index.html
+SH
+chmod +x "$ROOT/bin/smoke.sh"
+
+cat >"$ROOT/index.html" <<'HTML'
+<!doctype html>
+<title>Evening Reset</title>
+<main>Evening Reset</main>
+HTML
+
+python3 "$SCRIPT" init --root "$ROOT" >/dev/null
+python3 "$SCRIPT" test-hub-add \
+  --root "$ROOT" \
+  --title "Loose bad-case smoke" \
+  --command-text "./bin/smoke.sh" \
+  --run-policy every-dev-completion \
+  --test-status approved >/dev/null
+
+cat >"$ROOT/.codex/context/bad-cases.md" <<'MD'
+# Bad Case Register
+
+## Active Cases
+
+- BC-20260709-901 | 状态：resolved | 标签：日期切换, localStorage
+  - 现象：第二天打开页面时仍沿用前一天完成状态。
+  - 触发：localStorage 只保存勾选状态，不比较日期。
+  - 修复：保存日期并在跨天时重置当天状态。
+  - 验证：跨天打开后清单恢复待完成。
+- BC-20260709-902 | 状态: resolved | 标题: 部分完成刷新后进度丢失
+  - 现象: 勾选两项后刷新，进度回到 0。
+  - 触发: 只在最终完成时写入 localStorage。
+  - 修复: 每次勾选变化都保存进度。
+  - 验证: 刷新后保留已勾选项目。
+  - 标签: 日期切换, localStorage
+MD
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "loose-bc-agent" \
+  --summary "Subagent wrote loose bad-case register entries." >/tmp/context-guard-loose-bc.out
+
+grep -Fq "[context-guard] test hub: 1 passed, 0 failed, 0 blocked." /tmp/context-guard-loose-bc.out
+grep -Fq "created proposed chains: 1" /tmp/context-guard-loose-bc.out
+
+python3 - "$ROOT/.codex/context/test-hub/feature-chains.json" <<'PY'
+from pathlib import Path
+import json
+import sys
+
+data = json.loads(Path(sys.argv[1]).read_text(encoding="utf-8"))
+chains = data.get("chains", [])
+assert len(chains) == 1, chains
+chain = chains[0]
+assert chain["status"] == "proposed", chain
+assert chain["confirmation_required"] is True, chain
+nodes = chain["nodes"]
+assert len(nodes) == 2, nodes
+refs = [ref for node in nodes for ref in node.get("bad_cases", [])]
+assert refs == ["BC-20260709-901", "BC-20260709-902"], nodes
+titles = [node.get("title", "") for node in nodes]
+checks = [check for node in nodes for check in node.get("checks", [])]
+assert any("部分完成刷新后进度丢失" in item for item in titles), nodes
+assert any("刷新后保留已勾选项目" in item for item in checks), nodes
+PY

package/tests/BC-20260709-008.sh

@@ -0,0 +1,168 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(mktemp -d "${TMPDIR:-/tmp}/cg-risk-audit.XXXXXX")"
+trap 'rm -rf "$ROOT"' EXIT
+
+SCRIPT="/Users/bytedance/.agents/skills/context-guard/scripts/context_guard.py"
+
+python3 "$SCRIPT" init --root "$ROOT" >/dev/null
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "risk-audit-smoke" \
+  --summary "加入练习模式、关闭倒计时选项、失败复盘里的逐步回放，以及最高分重置确认；smoke 通过。" \
+  >"$ROOT/out-risk.txt"
+
+grep -q "risk audit: created bad-case candidate" "$ROOT/out-risk.txt"
+grep -q "risk-audit" "$ROOT/.codex/context/bad-cases.md"
+grep -q "游戏流程.*未验证" "$ROOT/.codex/context/bad-cases.md"
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "risk-audit-duplicate" \
+  --summary "加入练习模式、关闭倒计时选项、失败复盘里的逐步回放，以及最高分重置确认；smoke 通过。" \
+  >"$ROOT/out-duplicate.txt"
+
+grep -q "risk audit: no new bad-case candidate" "$ROOT/out-duplicate.txt"
+if [[ "$(grep -c '^### BC-' "$ROOT/.codex/context/bad-cases.md")" != "1" ]]; then
+  echo "duplicate risk-audit summary should not create a second similar bad case" >&2
+  exit 1
+fi
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "risk-audit-same-tags-new-step" \
+  --summary "继续加入关卡包切换和暂停恢复，失败后展示本轮序列复盘；smoke 通过。" \
+  >"$ROOT/out-same-tags-new-step.txt"
+
+grep -q "risk audit: created bad-case candidate" "$ROOT/out-same-tags-new-step.txt"
+if [[ "$(grep -c '^### BC-' "$ROOT/.codex/context/bad-cases.md")" != "2" ]]; then
+  echo "same-tag but different long-flow step should create another bad case candidate" >&2
+  exit 1
+fi
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "risk-audit-distinct" \
+  --summary "加入复制结果和导出 Markdown，空输入时显示校验提示；smoke 通过。" \
+  >"$ROOT/out-distinct.txt"
+
+grep -q "risk audit: created bad-case candidate" "$ROOT/out-distinct.txt"
+if [[ "$(grep -c '^### BC-' "$ROOT/.codex/context/bad-cases.md")" != "3" ]]; then
+  echo "distinct risk-audit summary should create a second bad case candidate" >&2
+  exit 1
+fi
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "risk-audit-edit-isolation" \
+  --summary "新增角色小队的章节计划，重生成单章时不影响其他章节，导出 Markdown 包含章节计划；smoke 通过。" \
+  >"$ROOT/out-edit-isolation.txt"
+
+grep -q "risk audit: created bad-case candidate" "$ROOT/out-edit-isolation.txt"
+grep -q "局部隔离" "$ROOT/.codex/context/bad-cases.md"
+if [[ "$(grep -c '^### BC-' "$ROOT/.codex/context/bad-cases.md")" != "4" ]]; then
+  echo "long-flow edit isolation should create a risk-audit bad case candidate" >&2
+  exit 1
+fi
+
+ROOT_LOW="$(mktemp -d "${TMPDIR:-/tmp}/cg-risk-audit-low.XXXXXX")"
+trap 'rm -rf "$ROOT" "$ROOT_LOW"' EXIT
+python3 "$SCRIPT" init --root "$ROOT_LOW" >/dev/null
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT_LOW" \
+  --agent-id "risk-audit-low" \
+  --summary "更新 README 文案并修正两个错别字；smoke 通过。" \
+  >"$ROOT_LOW/out-low.txt"
+
+grep -q "risk audit: no new bad-case candidate" "$ROOT_LOW/out-low.txt"
+if grep -q "risk-audit" "$ROOT_LOW/.codex/context/bad-cases.md"; then
+  echo "low-risk summary should not create risk-audit bad case" >&2
+  exit 1
+fi
+
+ROOT_LEGACY="$(mktemp -d "${TMPDIR:-/tmp}/cg-risk-audit-legacy-chain.XXXXXX")"
+trap 'rm -rf "$ROOT" "$ROOT_LOW" "$ROOT_LEGACY"' EXIT
+python3 "$SCRIPT" init --root "$ROOT_LEGACY" >/dev/null
+mkdir -p "$ROOT_LEGACY/.codex/context/test-hub"
+cat >"$ROOT_LEGACY/.codex/context/bad-cases.md" <<'MD'
+# Bad Case Register
+
+## Active Cases
+
+### BC-20260709-101: 状态切换风险未验证
+
+- Status: open
+- Scope: 状态切换
+- Tags: #状态流程 #回放 #risk-audit #subagent
+- Display summary: Subagent 本轮开发涉及状态切换和回放，但没有对应功能链节点。
+- Phenomenon: 练习模式和普通模式来回切换后状态可能串线。
+- Trigger / reproduction: 切换练习模式后退出，再继续普通模式。
+- Guard / verification: 运行游戏入口，检查练习模式和普通模式状态互不污染。
+
+### BC-20260709-102: 重置撤销风险未验证
+
+- Status: open
+- Scope: 状态切换
+- Tags: #状态流程 #重置 #risk-audit #subagent
+- Display summary: Subagent 本轮开发涉及重置撤销，但没有对应功能链节点。
+- Phenomenon: 重置确认取消后仍可能清空游戏进度。
+- Trigger / reproduction: 点击重置全部进度，再取消确认。
+- Guard / verification: 运行游戏入口，检查取消重置后进度仍保留。
+MD
+cat >"$ROOT_LEGACY/.codex/context/test-hub/feature-chains.json" <<'JSON'
+{
+  "version": 1,
+  "chains": [
+    {
+      "id": "FC-20260709-001",
+      "title": "Context Guard subagent completion risk audit",
+      "status": "proposed",
+      "run_policy": "every-dev-completion",
+      "entry": "触发内部审计",
+      "exit_check": "内部审计通过",
+      "command": "",
+      "timeout_seconds": 300,
+      "artifact_policy": "cleanup-on-pass",
+      "resource": "local",
+      "created": "2026-07-09",
+      "source": "feature-chain-auto-propose",
+      "auto_proposed": true,
+      "auto_group_key": "#risk-audit|#subagent",
+      "confirmation_required": true,
+      "nodes": [
+        {
+          "id": "FC-20260709-001-N1",
+          "title": "状态切换风险未验证",
+          "bad_cases": ["BC-20260709-101"],
+          "checks": ["检查练习模式和普通模式状态互不污染"]
+        },
+        {
+          "id": "FC-20260709-001-N2",
+          "title": "重置撤销风险未验证",
+          "bad_cases": ["BC-20260709-102"],
+          "checks": ["检查取消重置后进度仍保留"]
+        }
+      ]
+    }
+  ]
+}
+JSON
+
+python3 "$SCRIPT" feature-chain-auto-propose --root "$ROOT_LEGACY" >/tmp/risk-audit-legacy-refresh.out
+
+python3 - "$ROOT_LEGACY/.codex/context/test-hub/feature-chains.json" <<'PY'
+from pathlib import Path
+import json
+import sys
+
+data = json.loads(Path(sys.argv[1]).read_text(encoding="utf-8"))
+chain = data["chains"][0]
+assert chain["title"] == "状态切换", chain
+assert "risk-audit" not in chain.get("auto_group_key", ""), chain
+assert "subagent" not in chain.get("auto_group_key", ""), chain
+assert "#状态流程" in chain.get("auto_group_key", ""), chain
+assert chain["entry"] == "触发「状态切换」相关用户流程", chain
+assert chain["exit_check"] == "「状态切换」相关结果保持用户可见的正确状态", chain
+PY

package/tests/BC-20260710-001.sh

@@ -0,0 +1,61 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT="${CONTEXT_GUARD_SCRIPT:-$(cd "$(dirname "$0")/.." && pwd)/scripts/context_guard.py}"
+ROOT="$(mktemp -d "${TMPDIR:-/tmp}/context-guard-completion-evidence-XXXXXX")"
+trap 'rm -rf "$ROOT"' EXIT
+
+python3 "$SCRIPT" init --root "$ROOT" >/dev/null
+cat >"$ROOT/evidence.txt" <<'TXT'
+完成了跨空间移动和复制功能。
+CG_BAD_CASE: 跨空间移动会误删同名物品
+CG_PHENOMENON: 把一个物品移动到另一个空间时，源空间里名称相同但 ID 不同的物品也会被删除。
+CG_TRIGGER: 在源空间建立两个同名物品，只移动其中一个。
+CG_CAUSE: 删除源数据时使用名称匹配，而不是本次移动记录的物品 ID。
+CG_FIX: 移动完成后只按本次移动成功的 item.id 删除源物品。
+CG_VERIFICATION: 建立两个同名物品后移动其中一个，目标空间新增一项，源空间仍保留另一项。
+CG_SCOPE: 跨空间复制与移动
+TXT
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "evidence-agent" \
+  --summary "完成跨空间移动功能。" \
+  --evidence-file "$ROOT/evidence.txt" >"$ROOT/first.out"
+
+grep -Fq "completion evidence: archived concrete bad case(s)" "$ROOT/first.out"
+grep -Fq "跨空间移动会误删同名物品" "$ROOT/.codex/context/bad-cases.md"
+grep -Fq -- "- Status: resolved" "$ROOT/.codex/context/bad-cases.md"
+grep -Fq "只按本次移动成功的 item.id 删除源物品" "$ROOT/.codex/context/bad-cases.md"
+if grep -Fq "risk-audit" "$ROOT/.codex/context/bad-cases.md"; then
+  echo "concrete completion evidence must win over generic risk audit" >&2
+  exit 1
+fi
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$ROOT" \
+  --agent-id "evidence-agent" \
+  --summary "完成跨空间移动功能。" \
+  --evidence-file "$ROOT/evidence.txt" >"$ROOT/repeat.out"
+
+grep -Fq "subagent completion already processed" "$ROOT/repeat.out"
+[[ "$(grep -c '^### BC-' "$ROOT/.codex/context/bad-cases.md")" == "1" ]]
+[[ "$(grep -c 'Subagent completion handoff' "$ROOT/.codex/context/roadmap.md")" == "1" ]]
+
+NATURAL_ROOT="$(mktemp -d "${TMPDIR:-/tmp}/context-guard-natural-evidence-XXXXXX")"
+trap 'rm -rf "$ROOT" "$NATURAL_ROOT"' EXIT
+python3 "$SCRIPT" init --root "$NATURAL_ROOT" >/dev/null
+python3 "$SCRIPT" subagent-complete \
+  --root "$NATURAL_ROOT" \
+  --agent-id "natural-evidence-agent" \
+  --summary "修复了跨空间移动会按名称误删同名物品的问题，改为按 item.id 精确删除；验证两个同名物品只移动目标项。" \
+  >"$NATURAL_ROOT/out.txt"
+
+grep -Fq "completion evidence: archived concrete bad case(s)" "$NATURAL_ROOT/out.txt"
+grep -Fq "跨空间移动会按名称误删同名物品" "$NATURAL_ROOT/.codex/context/bad-cases.md"
+if grep -Fq "risk-audit" "$NATURAL_ROOT/.codex/context/bad-cases.md"; then
+  echo "natural concrete repair evidence must not become generic risk audit" >&2
+  exit 1
+fi
+
+echo "BC-20260710-001 passed"

package/tests/BC-20260710-002.sh

@@ -0,0 +1,111 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
+SCRIPT="${CONTEXT_GUARD_SCRIPT:-$ROOT_DIR/scripts/context_guard.py}"
+HOOK="${CONTEXT_GUARD_HOOK:-$ROOT_DIR/scripts/context_guard_hook.py}"
+BASE="$(mktemp -d "${TMPDIR:-/tmp}/context-guard-subagent-root-XXXXXX")"
+trap 'rm -rf "$BASE"' EXIT
+CONTROL="$BASE/control"
+CHILD="$BASE/child-product"
+mkdir -p "$CONTROL" "$CHILD"
+CONTROL="$(realpath "$CONTROL")"
+CHILD="$(realpath "$CHILD")"
+
+cat >"$CHILD/verify.sh" <<'SH'
+#!/usr/bin/env bash
+set -euo pipefail
+test -f product.txt
+grep -Fq "ready" product.txt
+SH
+chmod +x "$CHILD/verify.sh"
+printf 'ready\n' >"$CHILD/product.txt"
+
+python3 "$SCRIPT" subagent-register \
+  --root "$CONTROL" \
+  --agent-id "root-agent" \
+  --project-root "$CHILD" \
+  --task "完善本地商品流程" >"$BASE/register.out"
+
+grep -Fq "subagent project root: $CHILD" "$BASE/register.out"
+python3 "$SCRIPT" test-hub-add \
+  --root "$CHILD" \
+  --title "商品流程" \
+  --command-text "./verify.sh" \
+  --run-policy every-dev-completion \
+  --test-status approved >/dev/null
+
+cat >"$BASE/evidence.txt" <<'TXT'
+CG_BAD_CASE: 商品状态切换后没有保存
+CG_PHENOMENON: 切换商品状态并刷新后，页面恢复旧状态。
+CG_TRIGGER: 修改状态后刷新页面。
+CG_CAUSE: 状态切换没有写入本地存储。
+CG_FIX: 状态变更后立即持久化。
+CG_VERIFICATION: 修改状态后刷新，仍显示新状态。
+CG_SCOPE: 商品状态管理
+TXT
+
+python3 - "$CONTROL" "$BASE/evidence.txt" <<'PY' | \
+  (cd "$CONTROL" && python3 "$HOOK" subagent-stop) >"$BASE/hook.out" 2>"$BASE/hook.err"
+from pathlib import Path
+import json
+import sys
+
+print(json.dumps({
+    "agent_id": "root-agent",
+    "cwd": sys.argv[1],
+    "last_assistant_message": Path(sys.argv[2]).read_text(encoding="utf-8"),
+}, ensure_ascii=False))
+PY
+
+grep -Fq "project root: $CHILD" "$BASE/hook.err"
+grep -Fq "test hub: 1 passed, 0 failed, 0 blocked" "$BASE/hook.err"
+test -f "$CHILD/.codex/context/test-hub/last-run.json"
+grep -Fq "completion evidence: archived concrete bad case(s)" "$BASE/hook.err"
+
+python3 "$SCRIPT" subagent-complete \
+  --root "$CONTROL" \
+  --agent-id "root-agent" \
+  --evidence-file "$BASE/evidence.txt" >"$BASE/complete.out"
+
+grep -Fq "subagent project root: $CHILD" "$BASE/complete.out"
+grep -Fq "subagent completion already processed" "$BASE/complete.out"
+grep -Fq "商品状态切换后没有保存" "$CHILD/.codex/context/bad-cases.md"
+grep -Fq "Subagent completion handoff" "$CHILD/.codex/context/roadmap.md"
+[[ "$(grep -c 'Subagent completion handoff' "$CHILD/.codex/context/roadmap.md")" == "1" ]]
+[[ "$(grep -c '^### BC-' "$CHILD/.codex/context/bad-cases.md")" == "1" ]]
+if grep -Fq "Subagent completion handoff" "$CONTROL/.codex/context/roadmap.md"; then
+  echo "completion checkpoint leaked into the control workspace" >&2
+  exit 1
+fi
+
+python3 - "$CONTROL/.codex/context/subagents/assignments.json" "$CHILD" <<'PY'
+from pathlib import Path
+import json
+import sys
+
+data = json.loads(Path(sys.argv[1]).read_text(encoding="utf-8"))
+item = data["assignments"]["root-agent"]
+assert item["project_root"] == sys.argv[2], item
+assert item["status"] == "completed", item
+assert item["last_test_code"] == 0, item
+PY
+
+python3 "$SCRIPT" subagent-register \
+  --root "$CONTROL" \
+  --agent-id "root-agent" \
+  --project-root "$CHILD" \
+  --task "继续下一轮商品流程" >/dev/null
+
+python3 - "$CONTROL/.codex/context/subagents/assignments.json" <<'PY'
+from pathlib import Path
+import json
+import sys
+
+item = json.loads(Path(sys.argv[1]).read_text(encoding="utf-8"))["assignments"]["root-agent"]
+assert item["status"] == "running", item
+assert item["last_test_code"] == 0, item
+assert item.get("last_completion_fingerprint"), item
+PY
+
+echo "BC-20260710-002 passed"

package/tests/npm-install-smoke.sh

@@ -0,0 +1,53 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(mktemp -d "${TMPDIR:-/tmp}/context-guard-npm-install-XXXXXX")"
+ROOT="$(cd "$ROOT" && pwd)"
+trap 'rm -rf "$ROOT"' EXIT
+NODE_BIN="${NODE:-node}"
+
+SKILL_COUNT="$(find . -path './.git' -prune -o -name SKILL.md -print | wc -l | tr -d '[:space:]')"
+test "$SKILL_COUNT" = "1"
+
+"$NODE_BIN" bin/context-guard-skill.js --help >/dev/null
+"$NODE_BIN" bin/context-guard-skill.js install --target "$ROOT/skill" >/dev/null
+
+test -f "$ROOT/skill/SKILL.md"
+test -f "$ROOT/skill/scripts/context_guard.py"
+test -f "$ROOT/skill/README.zh-CN.md"
+INSTALLED_SKILL_COUNT="$(find "$ROOT/skill" -name SKILL.md -print | wc -l | tr -d '[:space:]')"
+test "$INSTALLED_SKILL_COUNT" = "1"
+
+CODEX_HOME="$ROOT/codex-home" "$NODE_BIN" bin/context-guard-skill.js install >/dev/null
+test -f "$ROOT/codex-home/skills/context-guard/SKILL.md"
+
+npm_config_global=true CODEX_HOME="$ROOT/global-codex-home" "$NODE_BIN" bin/postinstall.js >/dev/null
+test -f "$ROOT/global-codex-home/skills/context-guard/SKILL.md"
+
+PROJECT="$ROOT/project"
+"$NODE_BIN" bin/context-guard-skill.js init --root "$PROJECT" >/dev/null
+test -f "$PROJECT/.codex/context/index.md"
+
+HOOKS="$ROOT/hooks.json"
+CONFIG="$ROOT/config.toml"
+cat >"$CONFIG" <<'EOF'
+[features]
+codex_hooks = true
+multi_agent = true
+
+[model]
+name = "test"
+EOF
+"$NODE_BIN" bin/context-guard-skill.js install --target "$ROOT/skill-with-hooks" --with-hooks --hooks-target "$HOOKS" >/dev/null
+grep -Fq "$ROOT/skill-with-hooks/scripts/context_guard_hook.py" "$HOOKS"
+grep -Fq '"SubagentStart"' "$HOOKS"
+grep -Fq '"SubagentStop"' "$HOOKS"
+grep -Fq 'hooks = true' "$CONFIG"
+grep -Fq 'multi_agent = true' "$CONFIG"
+grep -Fq 'name = "test"' "$CONFIG"
+! grep -Fq 'codex_hooks' "$CONFIG"
+
+EMPTY_HOME="$ROOT/empty-codex-home"
+CODEX_HOME="$EMPTY_HOME" "$NODE_BIN" bin/context-guard-skill.js install --with-hooks >/dev/null
+grep -Fq '[features]' "$EMPTY_HOME/config.toml"
+grep -Fq 'hooks = true' "$EMPTY_HOME/config.toml"