sm-batch-refresh

SM Batch Refresh

This skill is used tobatch-maintain the coverage company pool— the highest-frequency but most tedious work for analyst teams.

Mandatory process (v0.3 hard constraints)

⛔ Before any analysis output, you must strictly execute the 5-step pre-start process of ../../core/preamble.md pre-start process in 5 steps

⛔ Before any output is completed, you must strictly execute the 6-step post-completion process of ../../core/postamble.md post-completion process in 6 steps

Output archiving follows the ../../core/output-archive.md naming convention Output acceptance is self-checked item by item according to the ../../core/acceptance.md checklist

Skipping any step is considered an incomplete task.

Batch Refresh special note: the [Preflight] in preamble Step 4 must list the target list + dimensions + estimated number of tool calls for this refresh, ensuring the user knows this is a large-batch task.

Applicable scenarios

Every Monday morning: refresh quotes, this week's announcements, and this week's news for the entire coverage pool
The first working day of each month: refresh financial metrics + shareholder structure for all companies
Before each quarterly earnings season: batch pre-fetch earnings disclosure calendars
Ad hoc full scan: batch-assess impact after industry events ("US restrictions updated, scan the AI compute chain")

Core tasks

Read the coverage pool list: pull all tickers from {coverage_root}/INDEX.md Filter scope
Filter scope: filter by user specification (full / sub-sector / individual)
Execute one by one: call the corresponding data-fetching tool for each ticker
Automatic archiving: write each company's update to {coverage_root}/{ticker}_{name}/data/{YYYY-MM-DD}-refresh.md
Generate overall report: summarize success / failure / significant changes

Output format

Overall summary

# Batch Refresh · {YYYY-MM-DD}

**范围**：{全量 82 家 / 光模块子赛道 22 家 / ...}
**维度**：{行情+公告+新闻 / 全部财务字段 / ...}
**用时**：{X 分钟}

## 摘要
- ✅ 成功更新：N 家
- ⚠️ 部分成功：M 家
- ❌ 失败：K 家

## 显著变化（关注点）

| Ticker | 名称 | 变化类型 | 详情 |
|---|---|---|---|
| 688256 | 寒武纪 | 重大公告 | 新增订单合同 (财报披露) |
| 688981 | 中芯国际 | 财务异常 | Q4 毛利率超预期 (财报披露) |

## 失败清单

| Ticker | 名称 | 失败原因 |
|---|---|---|
| {ticker} | {name} | iFind 超时 / 数据未披露 |

Update details for each target (written to the data/ directory)

# {ticker} {name} · Refresh {YYYY-MM-DD}

## 行情快照
- 收盘价 / 周涨跌 / 月涨跌 / YTD (财报披露-iFind)

## 财务最新
- 最近一期 PE / PB / PS / ROE (财报披露-iFind)
- 营收 / 净利 / 毛利率 同比 (财报披露-iFind)

## 股东变化
- 十大股东最新一期 (财报披露-iFind)
- 与上一期对比 (合理推演)

## 本周公告 / 新闻
- {公告 1} (财报披露)
- {新闻 1} (市场共识)

## 自动告警
- ⚠️ 任何"显著变化"必须在这里标出（毛利率突变、股东大幅减持、重大公告等）

## 仍需补的资料
- {缺失项}

Best practices for batch execution

Throttling

At least 200ms interval between each ticker to avoid triggering data source rate limits
iFind MCP supports at most 100 queries per session; split into batches if exceeded

Failure retry

Any ticker failure is immediately recorded in the failure list,do not interrupt the entire batch task
Report failures together after all runs are complete

Incremental update

Compare against the snapshot from the last refresh,only output fields that have changed
avoid rewriting all data every time

Task persistence

Long batch tasks must support interruption and resumption
Record "completed N/total M" in active-tasks.md, and continue from the breakpoint

Collaboration with other skills

Trigger deep research: if a "significant change" is detected for a target,automaticallytrigger sm-catalyst-monitor single-point analysis
Trigger red-team review: if a target rises more than 20% this week,automatically suggestusing sm-red-team
Trigger thesis update: If the financial data contradicts the assumptions in the previous thesis,proactively remindthe user to run sm-thesis update

Output acceptance (in addition to the general checklist)

The list of tickers refreshed this time has been listed
Each ticker has a "updated fields" summary
Failed tickers clearly list the reasons
Significant changes have been flagged and trigger follow-up skill suggestions

About this skill