dsh-industry-research plugin
A DeepSeek Harness plugin for industry research: four tools and two skills, every figure carries a source and a date
- 219 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The plugin fetches public sources over the network and saves page snapshots to disk
Install
Manual install
dsh plugin --profile demo add dsh-industry-researchUse your profile name and restart the profile afterwards. A git channel with a pinned commit is also available.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The plugin fetches public sources over the network and saves page snapshots to disk
This is third-party code. Review the repository files before installing.
What it does
dsh-industry-research turns an industry study into a replayable chain for DeepSeek Harness. industry_map builds a value-chain map: upstream, midstream and downstream nodes, edges and metric slots. A metric with a value must carry a sourceRef, and an empty slot counts as an explicit gap. industry_track follows public sources on policy and news through the built-in web access, filters them with allow and block host lists, saves page snapshots with a SHA-256 hash and appends a de-duplicated timeline.jsonl feed. company_scan reads workspace files (md, txt, csv, tsv, json) and produces short company cards with hashes, outlines and numeric candidates. industry_report assembles the map, timeline and cards into a verifiable report and checks the delivery contract before output: unfilled placeholders and references to missing evidence cause an error. Every artifact write goes into a versions.jsonl ledger with a hash. Two skills, industry-research-method and company-research-method, define the methodology, and the product uses public sources only and supports research without replacing the conclusion.
Who it is for. Analysts and researchers who run industry and company reviews in DeepSeek Harness and want every figure tied to a source and a date.
Good fit when
- An industry chain map is needed with source references for each metric
- Policy and news on a topic must be collected as a feed with page snapshots and hashes
- A set of company files must become a short card with an outline and numbers
- The final report must pass checks: no placeholders and no claims without evidence
Not a fit when
- You need investment recommendations or a stock price valuation: the plugin only prepares material and gives no advice
- Sources are closed or in PDF: version 1 does not parse PDF files
- You work outside DeepSeek Harness: the plugin targets its environment
Example request
Load the industry-research-method skill and research the industry: first industry_map, then industry_track on policy, then industry_reportLimitations
The plugin works only in DeepSeek Harness and installs with dsh plugin add; restart the profile after installing. PDF files are not parsed in version 1. Source tracking needs network access, and offline mode keeps only local data. It is a research support tool: it publishes no conclusions and does not replace the analyst. Decisions on numbers and interpretation stay with a person.
How to disable. Run dsh plugin --profile <name> remove dsh-industry-research or delete the plugin line from the profile patch. Created artifacts live in the industry-research directory of the workspace.
FAQ
Which tools does the plugin include?
Four: industry_map, industry_track, company_scan and industry_report. Two methodology skills, industry-research-method and company-research-method, load on request.
Why sourceRef?
Every metric with a value must reference a source and carry a date. A metric without a source fails validation and an empty slot is marked as an explicit gap that the report lists separately.
How do I install it?
With dsh plugin --profile <name> add dsh-industry-research (the npm channel) or from git with a pinned commit. Restart the profile after installing and check dsh --dump-config.
Can it work offline?
Yes, offline: true disables network access and keeps only local workspace data. By default access is on and limited with allow and block host lists.
How is artifact history kept?
Each map, card and report write appends a line to versions.jsonl with the path, SHA-256 hash, time and change type. Hashes are verified on read and a mismatch raises an error.
Discussion
No questions or reviews yet.
Related
PluginsDeepSeek Harness plugin: a research report where each claim binds to a source snapshot and a SHA-256 seal can be checked offline
A pack of 14 skills for public-data research in Chinese: a lead skill runs the project and 13 specialists cover market, competitors and valuation
Auto-Empirical Research Skills
Auto-Empirical Research Skills (AERS)
A plugin marketplace for empirical social science research: causal inference, econometrics and manuscript prep for top journals
Excel analysis skill that profiles a sheet, cleans it, reconciles totals against the sheet's own sums and then releases numbers.
The Dex plugin: an agent explores a database, builds dbt models and a semantic layer and tracks schema drift under cost control
Step-by-step causal inference: 14 skills pick a method, check assumptions, write R or Python code and stress-test the result.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| dsh-industry-research (this card) | no | 219 | |
| Hehe Industry Research Skill Pack | no | 326 | a skill pack for industry analysis, while the plugin has four tools and a hashed version ledger |
| DeepSeek Harness | yes | 280,265 | the shell itself, while the plugin adds industry methodology and a verifiable report to it |
README in short
The README, in Chinese with versions in five languages, covers what the plugin is, a three-step install, configuration, the four tools and two skills with call examples, the artifact version ledger, FAQ and privacy. A catalog in the DSH Desktop marketplace is offered for discovering plugins.
README badge
Original SKILL.md
---
name: industry-research-method
description: '行业研究方法论：产业链上中下游拆分法、供需分析框架、政策跟踪口径、数据口径纪律（每个数字必须有来源与 asOf）、缺口声明纪律（宁标缺口不编造）。做行业梳理、产业链建图、政策动态跟踪或行业报告撰写时使用；只查单一公司时用 company-research-method。'
whenToUse: '用户要求研究一个行业（梳理产业链、跟踪政策动态、写行业研究报告）时使用本技能；配合 industry_map / industry_track / industry_report 工具落盘可核查产物。只做公司层面研究时改用 company-research-method。'
metadata:
pack: dsh-industry-research
version: '0.1.0'
---
# 行业研究方法论（industry-research-method）
本技能规范一次行业研究的全流程，产出**每个数字都能回溯到来源**的可核查产物。合规底线：**仅供研究，不构成投资建议**；不接交易、不做实时告警、不使用需登录/付费/绕反爬的数据源。
## 工具配合关系
| 阶段 | 工具 | 产物 |
|---|---|---|
| 建图 | `industry_map` | `<industryRoot>/<行业>/chain.json`（校验过的产业链结构图） |
| 跟踪 | `industry_track` | `<industryRoot>/<行业>/timeline.jsonl`（政策/动态时间线） |
| 成文 | `industry_report` | `reports/<时间戳>/report.md` + `manifest.json`（引擎缺席时为 builtin-fallback） |
工作区默认根为 `industry-research/`（可用 `industryRoot` 配置）。方法论细节见 `references/frameworks.md`（供需框架与常见拆分模板），需要时再用 read 工具加载。
## 阶段 1：产业链拆分（industry_map）
1. 先调用 `industry_map({ industry })`（不带 chain）读取当前图与已登记来源；有用户笔记/文件时经 `seed`/`seedFiles` 登记为来源（得 ref，如 `S1`）。
2. 按上/中/下游拆分撰写 ChainMap：
- **上游**：原材料、关键投入品、设备与基础设施。
- **中游**：生产制造、核心工艺环节、主要参与者类型。
- **下游**：分销渠道、终端应用、消费者/客户类型。
- 边表示投入→产出关系，`note` 写明关系性质（如「原料供应」「渠道分销」）。
3. 每个节点挂指标槽位：**有 `value` 必须带 `sourceRef`**（sources.json 的 ref、URL 或工作区路径）并尽量带 `asOf`；暂时没有来源的指标**只写 `key` 不写 `value`**——这就是显式待补槽位，禁止填估计值。
4. 带 `chain` 再次调用 `industry_map`；校验失败（悬空边、无来源数值、非法 tier）按返回的问题清单修正后重试。
5. 工具返回的 `gaps` 是后续工作的清单，不是失败：逐项决定是补充来源还是保留待补。
## 阶段 2：政策与动态跟踪（industry_track）
1. 默认主题为「<行业> 行业 政策」与「<行业> 行业 动态 要闻」；用户给定点时经 `topics` 传入。
2. 条目纪律：每条带来源 URL、发布日期（来源无日期则为空，**不得编造日期**）、摘要与快照哈希；抓取失败的条目以纯引用形式记录并在 `note` 中说明。
3. 来源治理：部署方可用 `sourceAllowlist`/`sourceBlocklist` 收口（如只信 `gov.cn`）；被拦截与去重的计数在结果中如实报告。
4. `ctx.web` 未挂载或 `offline: true` 时工具会响亮失败——如实转告用户需要挂载 web provider 或关闭 offline，**不得**用记忆代替检索结果充数。
## 阶段 3：报告成文（industry_report）
1. 默认自动草稿即可成文；需要定制叙述结构时用 `draft` 撰写 sections + claims：