SuperAPI AI Directory

reg-feed-watcher

Data & analytics

View repository

Updated 10-15 20:29

About this skill

/reg-feed-watcher
  1. Read ~/.claude/plugins/config/claude-for-legal-zh/regulatory-legal/CLAUDE.md → monitoring list, importance threshold, dynamic source configuration.
  2. Use the following workflow.
  3. Pull each dynamic source. Filter by importance.
  4. Output: new items, classified by importance level.

Purpose

Pull dynamic sources. Filter by importance. Output what remains. The filter is where the value lies—unfiltered dynamic sources are noise.

Load context

~/.claude/plugins/config/claude-for-legal-zh/regulatory-legal/CLAUDE.md → monitoring list, importance threshold, dynamic source configuration, summary output path (if set).

Workflow

Step 0: Coverage check (before pulling)

Before pulling dynamic sources, compare the monitoring list and dynamic source configuration against the regulatory agency categories that need to be covered:

  • Which categories of regulatory agencies the user focuses on based on the monitoring list (State Council ministries/affiliated institutions/provincial level)
  • Which categories have zero or very few sources configured

If there is an obvious gap—for example, the user's monitoring list includes the "financial regulation" category but the dynamic sources only include "People's Bank of China" and lack "National Financial Regulatory Administration" and "China Securities Regulatory Commission"—prompt once at the top of the summary:

Coverage gap alert: Your monitoring list includes [category], but only [N] dynamic sources are configured. Suggest adding dynamic sources for [agency names]. Do you want suggestions for additions? Run /regulatory-legal:cold-start-interview --redo to update, or directly edit ~/.claude/plugins/config/claude-for-legal-zh/regulatory-legal/CLAUDE.md。

Do not repeatedly prompt about the same gap—if the user has clearly said "do not focus on a certain regulatory agency for now," respect that decision and record it in CLAUDE.md so it remains effective.

Step 1: Pull

Pull from all configured dynamic source tiers. Every installation has Tier 1. Tier 2 and Tier 3 are additional—use them if configured, skip them if not configured.

Tier 1—free dynamic sources (always active)

For each regulatory agency in the monitoring list:

Regulatory agency monitoring reference:

Regulatory agency Monitoring content Source
Cyberspace Administration of China (CAC) Personal information protection, data security, algorithm governance, content management cac.gov.cn
State Administration for Market Regulation Antitrust, anti-unfair competition, advertising, consumer protection samr.gov.cn
Ministry of Industry and Information Technology Telecommunications, internet information services, data security miit.gov.cn
Ministry of Science and Technology Science and technology ethics, artificial intelligence most.gov.cn
National Data Administration Data elements, data circulation, data infrastructure Government websites
Supreme People's Court Judicial interpretations, guiding cases court.gov.cn

Tier 2 — Paid dynamic sources (if configured)

  • Legal database MCPs (e.g., Yuandian, etc.): Query updates since the last check date, filtered by the regulators in the watchlist.
  • Other professional legal research tool MCPs: Same as above.

Cross-tier deduplication — the same document may appear in multiple sources. Use the most informative source to enrich the output.

No silent filling. If a dynamic source pull returns little or no results for a regulator in the watchlist, report what was found and stop. Do not fill the gaps with web search or model knowledge without asking. State: "The dynamic source check returned [N] items from [the regulator(s) hit]. Coverage for [regulator/topic] appears thin. Options: (1) expand the date window, (2) try a different dynamic source or MCP, (3) search the web — results will be marked [联网检索 — 需复核], and should be verified with the issuing authority before reliance, (4) stop here. Which would you like?" It is for the lawyer to decide whether to accept lower-confidence information sources.

Source provenance. Mark the source of every citation and regulatory item:[中国政府网]、[<监管机构>官网]、[元典MCP]、[法律数据库], or the specific MCP tool name (for items retrieved via a connector);[联网检索 — 需复核](for items retrieved via web search);[模型知识 — 需验证](for items surfacing from model training data);[用户提供](for manually pasted items). Items marked 需验证 carry a higher fabrication risk than tool-retrieved items and should be checked first. Never strip or collapse the labels.

Secondary sources. Some sources (legal tech media, law firm commentary, industry association briefings, academic institution reports, etc.) report on primary regulatory actions but are not primary sources. Any item pulled from such a source should be marked [次级来源] along with the source name label. In the summary, when a secondary-source item describes a regulator action, add the note: "→ Trace back to the primary source: [if the regulator link is known, otherwise 'look it up on <regulator>.gov.cn before relying']". Do not classify an item as "always important" based solely on the strength of a secondary source — downgrade it by one tier until a primary source is found.

Tier 3 — Manual entry

If the user pastes regulatory text or a summary rather than triggering from a scheduled dynamic source check: treat the pasted content as a single item, skip to Step 2 for classification, and record the source as "manual entry". No dynamic source pull is needed. This path works regardless of subscription status.

Record the check timestamp after pulling. The next scheduled run starts from this point in time.

Step 2: Classification

Each item is assigned an importance tier based on ~/.claude/plugins/config/claude-for-legal-zh/regulatory-legal/CLAUDE.md :

Item type Matches threshold
Formally issued administrative regulations/departmental rules Usually "always important"
Draft for comment Usually "worth reviewing" — and always record the comment deadline
Pre-comment/survey notice Worth reviewing — used forstrategyRather than compliance—no specific requirements have been imposed yet, but it signals direction and has a real feedback deadline. Record the deadline. Transmit only as preliminary analysis to /regulatory-legal:policy-diff, not as a gap analysis for gap elimination
Regulatory enforcement actions/administrative penalties Industry match→important; relevant practice match→worth reviewing; neither matches→for reference only or skip
Regulatory guidelines/guidance opinions Worth reviewing
Leadership speeches/policy briefings For reference only or skip based on threshold
Settlements/rectification As appropriate—novel theory or numerous cases→worth reviewing; routine→skip

Step 3: Enrich

For each item above the "for reference only" tier:

  • One-line summary (what changed)
  • Why it may be relevant here (relevance hook—"this concerns the [practice] you are working on")
  • Source link
  • Effective date or comment deadline (if applicable)

Do not summarize "for reference only" items one by one—just count them.

Output

By default, the summary is output in the conversation.When the output contains one or more items above "for reference only", also write to a shareable file, unless the user's CLAUDE.md explicitly sets 摘要输出 → 仅对话。

[工作成果头 — 按照插件配置 ## 输出 — 根据角色有所不同;见 `## 谁在使用此工具`]

## 法规动态检查 — [日期]

**期间:** [上次检查] 至 [现在]
**已检查的动态源:** [列出活跃的层级——如"中国政府网、网信办官网、元典MCP"]
**发现事项:** [N] 总计

### 要点

[N个差距需要在[日期]前采取行动 — 前3项:X、Y、Z]

### 🔴 始终重要

**[监管机构] — [标题]**
[一行摘要]。[关联性钩子]。于[日期]生效。
[链接]
→ 建议:对[可能受影响政策]运行政策差异分析

[对每项重复]

### 🟡 值得审阅

**[监管机构] — [标题]**
[一行]。[关联性]。[如有截止日期]。
[链接]

[征求意见稿:如意见征集跟踪已启用,包含"💬 征求意见截止日期:[日期] — 决策待定"]

[重复]

### 📝 仅供参考

[N] 项 — [可展开的标题+链接列表,无摘要]

---

**上次检查更新至:** [时间戳]
**意见征集中:** [N] 个征求意见稿有待决定——运行 /regulatory-legal:comments 审阅

---

**依赖前核实引注。** 此处的法规引注由AI生成,未经对照一手来源核实。在对上述任何法规、规章、指导意见或执法行动采取行动之前,通过法律研究工具(元典MCP、你的律所研究平台或发布机构的官方网站)确认其准确性、生效日期和当前状态。AI生成的法规引注有时是虚构、引用错误或过时的。每个事项上的来源标签(例如 `[中国政府网]`、`[联网检索 — 需复核]`)标明其出处;`需验证` 标签具有更高的编造风险,应首先检查。

Configuration-related fallbacks

  • Monitoring list is empty: Stop and state "The monitoring list in your configuration is empty. Without knowing which regulators to monitor, I cannot pull dynamic sources. Run /regulatory-legal:cold-start-interview --redo or edit ~/.claude/plugins/config/claude-for-legal-zh/regulatory-legal/CLAUDE.md and add at least one regulator."
  • Importance threshold is empty: Fall back to the default tier and add an explanatory note.
  • Dynamic source configuration is empty: Run only the China Government Network announcement check and add an explanatory note.

If nothing is above "for reference only": "All quiet. [N] items for reference only, no attention needed."

Handoff

  • Hand off to policy gap analysis: Any item that is "always important" and may affect policy → propose running a gap analysis.
  • Hand off to gap presentation: If the gap analysis finds gaps → track.
  • Hand off to comment period tracker: Any draft for comment classified above "skip" → if tracking is enabled, automatically record the comment deadline.

Closing

Close with the next-step decision tree specified in CLAUDE.md ## 输出 .


What this skill does not do

  • It does not read through every item. It classifies and enriches; deep reading is for items that pass through the filter.
  • It does not change the importance threshold. If the filter is wrong, edit CLAUDE.md.
  • It does not require a paid subscription. Free dynamic sources are the baseline; paid dynamic sources add depth.