caveman-evidence-review

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Review Caveman evidence

查看Caveman证据

Act as a read-only operator. Build conclusions from current Caveman data, not from repository guesses. Never start, approve, cancel, or roll back an experiment from this skill.

以只读操作员的身份开展工作。基于当前Caveman数据得出结论，而非基于仓库猜测。绝对不能通过此skill启动、批准、取消或回滚任何实验。

Hard rules

硬性规则

Keep these buckets separate:
- measured provider-complete list-price cost;
- ```
inferred
```
  daily headroom;
- ```
verified
```
  ledger savings;
- evidence cost. Never add or relabel them.
Do not fetch prompt, completion, tool, or artifact payloads unless the user explicitly asks for payload review. Metadata, spans, timing, models, token counts, status, and optimizer attribution are enough for the default review.
Scope every read to the project selected by Caveman context. Never supply an organization id.
Empty results are evidence of no current signal, not zero cost or zero risk.
Cite trace ids and exact time windows used. Do not claim a cause from an aggregate alone.

请将以下类别严格区分：
- 已计量的供应商完整标价成本；
- ```
inferred
```
  （推断的）每日可用空间；
- ```
verified
```
  （已验证的）账本节省金额；
- 证据成本。切勿对这些类别进行添加或重新标记。
除非用户明确要求查看负载内容，否则请勿获取提示词、完成内容、工具或工件负载。默认审查仅需元数据、跨度、计时、模型、令牌数量、状态以及优化器归因信息即可。
所有只读操作的范围均限定为Caveman上下文所选的项目。切勿提供组织ID。
空结果代表当前无信号，而非零成本或零风险。
引用所使用的trace ids和精确时间窗口。仅通过聚合数据不得断言因果关系。

Step 1 — Load context

步骤1 — 加载上下文

Prefer MCP:

text

caveman_context {}

CLI fallback:

bash

caveman cloud whoami
caveman cloud projects list

Stop if login or project selection is missing. Ask the user to run

caveman login

or select a project; never guess.

优先使用MCP：

text

caveman_context {}

CLI备选方案：

bash

caveman cloud whoami
caveman cloud projects list

若缺少登录信息或项目选择，请停止操作。请用户运行

caveman login

或选择一个项目；切勿猜测。

Step 2 — Establish baseline

步骤2 — 建立基准

Use

caveman_report

for:

```
overview
```
```
costs
```
```
score
```
```
workflows
```
```
verified_savings
```

Then use

caveman_plan

for ranked daily headroom. If question is narrow, skip unrelated reports. Read shortest set that can answer it.

CLI fallback:

bash

caveman cloud costs
caveman cloud score
caveman cloud plan --json

State report window and basis before interpreting direction.

使用

caveman_report

获取以下内容：

```
overview
```
（概览）
```
costs
```
（成本）
```
score
```
（分数）
```
workflows
```
（工作流）
```
verified_savings
```
（已验证节省金额）

然后使用

caveman_plan

获取按优先级排序的每日可用空间。若问题范围较窄，可跳过无关报告。读取能够回答问题的最少报告集合。

CLI备选方案：

bash

caveman cloud costs
caveman cloud score
caveman cloud plan --json

在解读趋势前，请说明报告的时间窗口和依据。

Step 3 — Test the leading explanation with traces

步骤3 — 用追踪记录验证主要解释

Use

caveman_trace_search

. Choose a bounded window and closed filters: workflow, agent, model, provider, error code, runtime mode, cache status, optimization id, status class, token/cost/latency bounds, compression, or monitor verdict.

Useful groupings:

```
workflow
```
— find jobs driving cost or failures;
```
model
```
— compare model mix;
```
session
```
— isolate retry or loop behavior;
ungrouped — identify exact traces.

Compare a suspect cohort with a control cohort or earlier bounded window. Do not infer causality from one expensive trace.

CLI fallback:

bash

caveman cloud traces search \
  --workflow <slug> \
  --from <RFC3339> \
  --to <RFC3339> \
  --sort total_cost_usd \
  --dir desc \
  --limit 25

使用

caveman_trace_search

。选择一个限定的时间窗口和筛选条件：工作流、Agent、模型、供应商、错误代码、运行时模式、缓存状态、优化ID、状态类别、令牌/成本/延迟范围、压缩或监控 verdict。

实用分组方式：

```
workflow
```
— 找出驱动成本或导致故障的任务；
```
model
```
— 对比模型组合；
```
session
```
— 隔离重试或循环行为；
未分组 — 识别具体的追踪记录。

将可疑群组与对照组或更早的限定时间窗口进行对比。不得仅从一条高成本追踪记录推断因果关系。

CLI备选方案：

bash

caveman cloud traces search \
  --workflow <slug> \
  --from <RFC3339> \
  --to <RFC3339> \
  --sort total_cost_usd \
  --dir desc \
  --limit 25

Step 4 — Inspect representative traces

步骤4 — 检查具有代表性的追踪记录

Call

caveman_trace_get

for a small number of high-signal trace ids. Inspect request and span metadata, latency, status, token counts, cache state, applied optimizers, and model route. Keep payload retrieval off.

CLI fallback:

bash

caveman cloud traces show <trace-id> --spans

调用

caveman_trace_get

获取少量高信号的trace ids。检查请求和跨度元数据、延迟、状态、令牌数量、缓存状态、已应用的优化器以及模型路由。请勿获取负载内容。

CLI备选方案：

bash

caveman cloud traces show <trace-id> --spans

Step 5 — Report

步骤5 — 生成报告

Use this shape:

text

undefined

使用以下格式：

text

undefined

Caveman evidence review

Caveman证据审查报告

Scope: <project> · <from> to <to> Measured cost: <value and basis> Verified savings: <ledger value, kept separate> Inferred headroom: <per-day band, kept separate>

Findings:

<finding> — <aggregate evidence> — traces <ids>
<finding> — <aggregate evidence> — traces <ids>

Unproven:

<plausible explanation lacking a control, trace, or eval>

Next read-only check:

<one bounded query>

Possible action:

<proposal only; use caveman-manage for read-only lifecycle review and safety gate>


If data is missing, name missing signal and stop at strongest supported
statement. Never turn a catalog subtotal into an invoice or an experiment result
into verified savings.

范围：<项目> · <开始时间> 至 <结束时间> 已计量成本：<数值及依据> 已验证节省金额：<账本数值，单独列出> 推断可用空间：<每日区间，单独列出>

发现：

<发现内容> — <聚合证据> — 追踪记录 <ID>
<发现内容> — <聚合证据> — 追踪记录 <ID>

未证实内容：

<缺乏对照组、追踪记录或评估的合理解释>

下一步只读检查：

<一个限定范围的查询>

可能的操作：

<仅为提议；如需只读生命周期审查和安全把关，请使用caveman-manage>


若数据缺失，请说明缺失的信号，并基于最有力的支持性陈述停止操作。切勿将分类小计当作发票，也切勿将实验结果当作已验证的节省金额。