Screen publicly announced debt targets by industry, region, and amount range, and see which notice-listed creditors are associated with the matching results.
安装与下载
1. ClawHub CLI
从源直接安装技能的最快方式。
npx clawhub@latest install debtop-debt-screening
2. 手动安装
将技能文件夹复制到以下位置之一
~/.openclaw/skills/ <project>/skills/ 优先级：工作区 > 本地 > 内置
3. 提示词安装
将此提示词复制到 OpenClaw 即可自动安装。
请帮我使用 Clawhub 安装 debtop-debt-screening。如果尚未安装 Clawhub，请先安装（npm i -g clawhub）。
更倾向于下载？
获取包含原生技能文件的ZIP压缩包。
什么是 Debt Target Screening？
Debt Target Screening helps users find and organize public debt transfer, disposal, and investment-invitation notices. It searches notice text using industry, asset-type, collateral, or location keywords, then filters the results and groups matching targets by the creditor or seller named in each notice. Openclaw Skills users can review target counts, principal totals, debtors involved, and links to the original notice details.
Amount and region filters are applied locally to the records scanned; the search API does not accept those filters directly. Results therefore disclose the total reported matches, records scanned, coverage rate, and date window. This is read-only research based on public notices—not a credit report or investment recommendation. A notice-listed creditor is not necessarily the debt’s current holder.
Debt Target Screening 应用场景
- Find public NPL or debt-disposal notices related to a specific industry, such as mining, within a requested amount range.
- Identify which notice-listed creditors are associated with debt targets in a particular region or asset category.
- Run an initial screen of public debt targets before reviewing selected notices in detail.
- Summarize the distribution of scanned targets by seller, amount, region, or notice type.
- Compare likely search terms when a broad keyword produces too many results or an exact term returns none.
- Use a more specialized Openclaw Skills workflow when the request is only for a keyword notice list, one creditor’s disposal list, or a reverse lookup from collateral clues.
Debt Target Screening 工作原理
- Check whether the debt MCP tools are already available. If they are, use them directly; do not create a guest identity unnecessarily.
- If tools are unavailable, make a lightweight endpoint check. Use the guest read-only route only when the MCP responds with an eligible 401 and the client cannot complete OAuth; do not create a guest for rate limits, tool errors, or connectivity failures.
- Translate the request into focused, single-term searches. Search terms can match notice titles, creditor or debtor names, and disclosed collateral text. Do not combine multiple terms in one keyword parameter.
- Scan matching notices in pages of up to 20, deduplicate by notice ID, and stay within a reasonable request budget. Narrow large result sets with more specific terms before scanning further.
- Apply region and amount filters to the records actually retrieved. Prefer the notice’s area field for region matching; use title or debtor text only as a clearly disclosed fallback. For amounts, use principal when available and fall back to total amount when necessary.
- Aggregate qualifying notices by the creditor named in the notice. Calculate target counts and principal totals, deduplicate debtor names, and sort by principal total. Optionally retrieve detail pages for selected notices or enterprise debt summaries.
- Present the target list and original detail URLs alongside the reported total, scanned count, coverage rate, and scanned date window. State any partial coverage and the public-notice limitations.
Debt Target Screening 配置指南
No separate registration, login, API key, or skill-specific configuration is required. The agent platform should use an existing Debtop MCP connection when its tools are available. If using the guest fallback, the runtime must be able to access HTTPS and run the bundled Bash and Python scripts; guest access is read-only and subject to shared rate limits. The endpoint and tokens are authorized only for agent.debtop.com.
The skill package includes reference documentation and helper scripts. From the skill directory, inspect the available resources with:
ls references scripts templates
Only use the guest-token helper when the MCP is not configured and the authentication decision permits guest fallback:
bash scripts/guest_token.sh
For a manually managed guest challenge, the bundled proof-of-work helper is invoked as follows:
python3 scripts/pow_solve.py <challenge> <difficulty>
When the MCP tools are available, the four read-only tools are search_debt_notice, get_debt_notice_detail, search_debt_enterprise, and get_debt_enterprise_debt_summary. The notice search accepts keyword, page, and page_size; it does not accept amount, region, or notice-type filters. Keep calls sequential and at least one second apart. The guest channel is limited to 1 QPS, 30 calls per minute, and 1,000 calls per day, with additional guest-creation quotas.
Debt Target Screening 数据架构与分类体系
The skill works with public notice and enterprise search results. It does not describe a local database or create a persistent archive; results are filtered and aggregated for the response. The report should preserve the service-returned detail_url exactly as provided.
| Data group | Fields and handling |
|---|---|
| Notice identity and source | notice_id identifies a notice; detail_url links to its original detail page; notice_date supports date-window reporting. Deduplicate multi-keyword results by notice_id. |
| Notice classification | title, area, and notice text support relevance and region screening. area is a notice-region field, generally at province level, and may be missing; it does not necessarily identify collateral location. |
| Parties | creditor is the seller or creditor named in the notice, not confirmation of the current debt holder. debtor may contain multiple names; split comma-separated names before counting and deduplicating. |
| Amounts | Use principal_amount_yuan for filtering and aggregation when disclosed. If unavailable or zero, fall back to total_amount_yuan and label that basis. If neither is available, report the record in a separate amount-undisclosed group. Use numeric *_yuan fields for calculations and corresponding *_text fields for display. |
| Optional detail and enterprise data | Notice details may include collateral, guarantor, and original_creditor. Enterprise lookup uses enterprise_id and may return an enterprise debt summary. Use IDs only when supplied by search results. |
| Screening and coverage metadata | Report query terms, filters, API total, actual scanned count, coverage rate (scanned / total), and the scanned notice_date window. If total is capped at 10,000, describe it as a possible upper-bound estimate rather than a confirmed full count. |
| Report organization | Present a screening overview, coverage statement, creditor distribution, selected target list with original links, distribution statistics, and public-data caveats. Amount-range results apply only to the scanned window unless coverage is complete. |
Debt Target Screening 高级特性
- Full-text keyword matching can surface industry and asset clues from notice titles, party names, and disclosed collateral text.
- Multi-term screening uses separate searches followed by local deduplication and intersection or union, rather than unsupported multi-keyword API filters.
- Large result sets can be narrowed with specific synonyms, seller names, known debtor names, or separate asset-type searches before pagination.
- Local amount filtering supports principal-first logic, a labeled total-amount fallback, and a separate group for undisclosed amounts.
- Creditor-level aggregation calculates target counts, principal totals, debtor sets, and latest notice dates; selected records can be enriched with notice details or enterprise debt summaries.
- Coverage-aware reporting discloses scanned counts, date windows, partial coverage, and potentially capped totals so results are not presented as exhaustive when they are not.
- The guest access workflow supports read-only access without user sign-up when eligible, with token refresh, OAuth upgrade guidance, error handling, and rate-limit-aware sequential calls.
- Compliance safeguards require public-notice attribution, query time, original unmodified detail links, and clear statements that results are not credit reports or investment advice.
SKILL.md
加载中
Openclaw 相关技能
XPAI Camera Control
通过安全、具有状态的 MCP 服务器发现、查看和管理兼容的 LAN 摄像头，提供流传输、快照、PTZ 和摄像头事件工具。
LadybugDB openCypher
使用 openCypher 和经过精心范围界定的 Python 辅助工具设计、查询和验证 LadybugDB 属性图。
Contract Triage Lite
Contract Triage Lite 可离线检查自由职业者合同中的四类常见风险，标出相关条款和缺失的保护措施，且不会将文档文本发送到任何地方。
精选*
