Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Lightweight playbook distilled from AI Architecture to keep dual-engine memory (.ai_context) and manifest dispatcher with minimal overhead; use when bootstrapping or porting the pattern into Claude Skills format.
.claude/skills/aiskillstore-ai-architect-lite/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 28% | 0% |
> 精简版「AI Architecture」：保留 .ai_context 记忆协议与 manifest 驱动调度，服务轻量项目或快速移植场景。
.ai_context + slash-first 调度。PROJECT_ROOT = 当前仓库；TOOLKIT_PATH = 本技能所在目录（用于引用资源）。PROJECT_ROOT/.ai_context/03_ACTIVE_TASK.md；缺失则按 references/lite-protocol.md 模板创建或运行 python skills/project-init/entrypoint.py。03_ACTIVE_TASK.md 获取 Current Mission；若存在 01/02 则作为补充规则加载，不强制解析全部细节。.ai-manifest.json：优先使用 commands 中的 slash 触发；若无 cognitive_skills 配置，则保持纯命令模式。references/superpowers-lite.md，按其“迷你 TDD”“快速执行模板”走一遍。python scripts/plan_helper.py ...；② 填写/创建 .ai_context/03_ACTIVE_TASK.md；③ 用 scripts/append_log.py 记录第一次动作）。python scripts/plan_helper.py --goal \"...\" --steps \"...\" --validation \"...\" --out - 获取模板）；② 先写“会失败”的验证点（命令或检查点），再动手；③ 完成后按验证点复跑并记录结果。python scripts/append_log.py --note "..." --action "..." --changes "..." --outcome "..." --next "..."（随技能提供的最小脚本，默认在项目根执行并自动创建 .ai_context/03_ACTIVE_TASK.md）。若项目已有通用 skills/memory-sync/entrypoint.py，可直接复用；两者都不可用时在 03_ACTIVE_TASK.md 手动追加四段式记录。PROJECT_ROOT；执行前用中文说明修改/命令目的与影响，避免写入密钥。references/，按需加载；核心指令保持在本文件。嗨，我是 AI Architect Lite（目前只加载了我）。为了让你轻松上手，我们可以从这三步选一：
1) 想清楚要做什么：用“迷你计划”列出目标、3 步以内的行动、以及如何确认完成（好处：先定方向，少走弯路）。
2) 准备一本“任务笔记”：如果没有 .ai_context/03_ACTIVE_TASK.md，就按 references/lite-protocol.md 的格式建一个，记录当前任务和进展（好处：状态清晰，随时可查）。
3) 留下一条“开工日志”：用 scripts/append_log.py 记下你做的第一步和结果（好处：有迹可循，方便回顾或求助）。
告诉我你的目标，或者直接挑一条开始，我会陪你一步步完成。references/lite-protocol.md：.ai_context 最小规范、开发日志模板、slash-first 调度要点。references/superpowers-lite.md：超能力精简卡，涵盖迷你 TDD、系统化调试和透明协作的最小做法。scripts/append_log.py：为新手准备的最小日志脚本，自动建文件并按模板追加；可作为示例扩展。scripts/plan_helper.py：生成目标/约束/步骤/验证的极简计划模板，默认输出到终端，避免额外文件。assets/：保留占位，用于未来模板/示例；当前无需加载到上下文。markdown### [2025-11-21 10:00] Action: 初始化 lite 协议 - Changes: 创建 .ai_context/03_ACTIVE_TASK.md；检查 .ai-manifest.json - Outcome: 完成；未运行测试 - Next: 按当前任务执行并在每步后同步日志
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→pass | 17,514 | 13,545 | -23% | 1 | 1 | 0% | 2,091 | 2,787 | +33% | 0 | 0 | — |
case-01 | fail→pass | 18,185 | 14,983 | -18% | 1 | 1 | 0% | 1,975 | 2,680 | +36% | 0 | 0 | — |
case-02 | fail→pass | 18,118 | 18,982 | +5% | 1 | 1 | 0% | 2,467 | 3,905 | +58% | 0 | 0 | — |
case-04 | fail→pass | 14,988 | 13,985 | -7% | 1 | 1 | 0% | 2,422 | 2,658 | +10% | 0 | 0 | — |
case-05 | fail→pass | 23,188 | 11,519 | -50% | 1 | 1 | 0% | 1,727 | 2,203 | +28% | 0 | 0 | — |
case-06 | fail→pass | 13,154 | 11,140 | -15% | 1 | 1 | 0% | 2,063 | 2,182 | +6% | 0 | 0 | — |
case-07 | pass→pass | 23,892 | 14,981 | -37% | 1 | 1 | 0% | 2,741 | 2,802 | +2% | 0 | 0 | — |
case-08 | pass→pass | 11,501 | 9,866 | -14% | 1 | 1 | 0% | 1,683 | 1,880 | +12% | 0 | 0 | — |
case-09 | pass→pass | 22,447 | 19,398 | -14% | 1 | 1 | 0% | 2,525 | 2,727 | +8% | 0 | 0 | — |
case-10 | pass→pass | 15,295 | 8,395 | -45% | 1 | 1 | 0% | 1,561 | 1,629 | +4% | 0 | 0 | — |
case-11 | pass→pass | 17,529 | 16,882 | -4% | 1 | 1 | 0% | 2,012 | 2,750 | +37% | 0 | 0 | — |
case-12 | pass→pass | 13,594 | 7,654 | -44% | 1 | 1 | 0% | 1,713 | 2,395 | +40% | 0 | 0 | — |
case-13 | fail→pass | 9,180 | 10,261 | +12% | 1 | 1 | 0% | 1,524 | 2,097 | +38% | 0 | 0 | — |
case-14 | fail→pass | 14,068 | 9,665 | -31% | 1 | 1 | 0% | 2,190 | 1,926 | -12% | 0 | 0 | — |
case-15 | fail→pass | 14,484 | 7,682 | -47% | 1 | 1 | 0% | 2,221 | 1,650 | -26% | 0 | 0 | — |
case-16 | fail→pass | 18,330 | 3,918 | -79% | 1 | 1 | 0% | 2,094 | 1,775 | -15% | 0 | 0 | — |
case-17 | pass→pass | 12,500 | 2,623 | -79% | 1 | 1 | 0% | 921 | 1,531 | +66% | 0 | 0 | — |
case-18 | fail→pass | 12,193 | 12,477 | +2% | 1 | 1 | 0% | 1,861 | 2,303 | +24% | 0 | 0 | — |
case-19 | fail→pass | 15,138 | 9,290 | -39% | 1 | 1 | 0% | 2,430 | 2,651 | +9% | 0 | 0 | — |
case-20 | fail→pass | 21,430 | 8,539 | -60% | 1 | 1 | 0% | 2,577 | 2,633 | +2% | 0 | 0 | — |
case-21 | pass→pass | 18,560 | 13,458 | -27% | 1 | 1 | 0% | 1,994 | 2,502 | +25% | 0 | 0 | — |
case-22 | fail→fail | 25,860 | 20,127 | -22% | 1 | 1 | 0% | 4,247 | 3,762 | -11% | 0 | 0 | — |
case-23 | fail→fail | 32,126 | 24,258 | -24% | 1 | 1 | 0% | 5,301 | 5,823 | +10% | 0 | 0 | — |
case-24 | fail→fail | 32,853 | 15,580 | -53% | 1 | 1 | 0% | 4,826 | 2,909 | -40% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted. The headline lift of +54 percentage points is the difference between those two pass rates over the 24 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.