SkillFoxx

AI Native Jobhunt skill

AI Native 求职全案

Eight modules for an agent at each job search step: vacancy triage, resume, interviews and offer comparison

  • 105 ★ on the repository
  • Last commit
  • MIT
Skill

Extended access

Why this level

  • Works with the applicant's resume and personal data in a local workspace
  • The application module can trigger real submissions through a connected executor, and a platform may restrict the account
All reasons and checks

gilgameshcc/ai-native-jobhuntAuthor:gilgameshcc

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/ai-native

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Works with the applicant's resume and personal data in a local workspace
  • The application module can trigger real submissions through a connected executor, and a platform may restrict the account

This is third-party code. Review the repository files before installing.

What it does

AI Native Jobhunt is a job search method written for an agent to read (Claude Code, Codex, Cursor), with documentation in Chinese. It is a set of `SKILL.md` modules, each covering one moment of the search. The `00-workspace` module creates a local workspace with positioning, facts and ledger files. The `01-resume` module tailors a resume to a vacancy using verifiable facts only and exports it to PDF. The `jd-triage` submodule rates a vacancy against conditions you wrote yourself and assigns one of four levels. The `interview-strategic-assessment` module checks before an interview how sound the company is and prepares questions. The `02-apply` module sends applications through a judgment gate: with no decision about a vacancy, the application does not go out, and on anti-fraud signals from the platform the work stops. The set includes a `tools/apply_budget.py` script on the Python standard library to count applications. `04-interview` keeps a debrief card after an interview, `05-offer` compares offers and shows which assumption would flip the order, and `06-positioning` uses application statistics to say whether to change wording, thresholds or direction. Modules connect only through files and work on their own: you can run any of them by handing the agent its `SKILL.md` and a description of your situation.

Who it is for. People who search for a job with an agent and want it to invent no experience in the resume and to show why a vacancy fits or not, especially Chinese-speaking users.

Good fit when

  • Before applying you need to know if a vacancy deserves the time, by your hard conditions
  • The resume must be tailored to a vacancy without embellishing experience
  • An interview is tomorrow and you need questions for the company marked by whether they will answer
  • Two job offers and the current job must be compared without a single summary score
  • Applications get no replies and you need to know whether wording or direction is at fault

Not a fit when

  • You need a mass application script, since the set provides decision discipline and does not replace an application executor
  • You do not read Chinese: the documentation and examples are written in Chinese
  • You want ready integrations with job boards: the set has none
  • You need a ready resume template without a workspace and fact checking

Example request

Here is modules/03-company-analysis/jd-triage/SKILL.md. I will paste a vacancy below, rate it against my conditions: remote and direct work with the business

Limitations

  • The documentation, modules and examples are written in Chinese. The agent copes with them, but you read the method in Chinese.
  • The author admits a sample of one person, and all thresholds in the text are provisional and marked with a confidence level. Recalibrate them for yourself.
  • The `02-apply` module does not submit applications itself. Real actions on the site come from your already connected executor, and the set only decides whether to send an application.
  • The README does not describe installation: the generic `npx skills add gilgameshcc/ai-native-jobhunt` fits because the modules hold `SKILL.md` files, but the author does not give it.

How to disable. Delete the module folders from the agent's skills directory. The workspace created by `00-workspace` and its files stay in your folder: delete them by hand if you do not need them.

FAQ

How do I connect AI Native Jobhunt to an agent?

Install the set with `npx skills add gilgameshcc/ai-native-jobhunt` or hand the agent the needed `SKILL.md` from `modules/` and describe your situation. It is not tied to one agent; any agent that reads and writes local files will do.

Can I start somewhere other than 00-workspace?

Yes. No module requires it first: you can pass a vacancy straight to `jd-triage` or two offers to `05-offer`. A module names missing files itself and runs in a reduced mode.

Does the agent invent facts in the resume?

The `01-resume` module treats a fact pool as the only source of truth, and every fact traces back to a record. It optimizes wording and adds no experience.

How does the set differ from projects like career-ops?

The author calls those good end-to-end projects and covers the judgments before and after a run: hard exclusion conditions, company assessment before an interview, and offer ranking with flip points.

Can I use the set with another agent?

Yes. The method is written in Markdown and tied to no product: `SKILL.md` is just a file name convention, and you can hand the whole file to any agent. The workspace constraint file `AGENTS.md` can be renamed if needed.

Discussion

No questions or reviews yet.

Skills
Needs a VPN

A Claude Code workflow where the agent searches job postings, scores fit against your profile, drafts a CV and cover letter, and helps prepare for interviews

WorkflowExtended access71.5K

A skill set for resumes and job search: tailoring to a vacancy, ATS pass-through checks, cover letters and interview preparation

SkillMinimal access3.2K
Editors’ pick

Job search inside your coding agent: scores each listing against your CV from 1 to 5, tailors a PDF resume, drafts letters and tracks applications

WorkflowExtended access82.3K

A job-hunt skill bundle: JD decoding, template resumes, behavioral interview prep, offer comparison and negotiation

SkillExtended access572

JustHireMe is a local-first job-search desktop app that gives an agent a skill and a stdio MCP server to work with job data safely.

SkillExtended access2.3K

A Claude Code kit: a skill scores a job posting, a builder makes CVs and letters, and a daily scanner looks for new ads.

SkillExtended access76

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialRepository starsMain difference
AI Native Jobhunt (this card)no105
career-opsno82,343an end-to-end application pipeline with vacancy scoring; here the modules assess before and after a run
AI Job Searchno71,466a full job search flow with material generation and no hard rejection conditions

README in short

The README gives a table of your situation and the module to use, a diagram of links between the eight modules and a comparison with existing projects. At the end come the author's known limitations and collapsed sections on run order, moving to another agent and the structure of each `SKILL.md`.

README badge

Original SKILL.md

# SKILL · JD 结构化与定级（v0.1.2 骨架）

**把本文件整份交给你的 agent 即可**——不依赖任何特定 agent 产品，不需要注册命令、不需要放在特定目录、不需要装任何东西。`SKILL.md` 只是一个文件名约定，不是某一家的格式。连同一份 JD 原文一起贴给它，就能开始。

> ⚠️ **这一版是骨架**：给的是规则与标定方法，**不是一套调好的参数**。文中所有 `〔…〕` 都是**留空的参数**，需要你自己标定（方法见第三步）。完整定级模板排在 v0.3。

---

## 交付约定：易用性与完成度优先

Agent 负责整理并完成本次可做部分，交付可直接人工 review 的成品，不止给建议或空表。先读用户工作区的 `policies/`；用户可配置的执行偏好和参数集中在这里维护，已有求职条件仍以 `positioning.md` 为准。明确要求的调整直接写回并在 `policies/changes.md` 记录旧值、新值、理由与影响；Agent 建议标待确认。

用户已指定路径就直接创建本次所需目录；未指定时一次询问存放位置，同时继续整理材料。不得要求用户先手工建文件夹，不得把个人材料写进公开仓库。具体岗位建立或复用 `applications/日期-公司-岗位/README.md`，索引实际成品，初始化只补缺失文件；原始材料和历史审阅保留，策略、申请索引及衍生产物可按本次授权更新并记录变更。结束前写 `reviews/日期-任务.md`：成品路径、结论依据、采用策略、待确认项及影响、实际检查和下一步；重新读取文件核对后给可点击路径。缺材料时先完成不依赖它的部分并标部分完成；无文件写入能力须明确未落盘。可审阅不等于获准发送，外部动作仍按既有授权和模块边界。

交付前交叉核对报告、申请索引、台账和审阅卡：结论与下一步必须一致，已判不投的岗位只说明排除依据及证据或本人条件改变后可重评，不沿用继续推进的话术。内容重排后修正上下文指代；检查数量只填写工具实际统计值，未统计就列检查对象，不手写估数。发现历史审阅错误时另写勘误，保留原记录。

完整规则与模板见 [本地交付约定](../../../docs/delivery.md)；只拿到本 SKILL 时按以上约定建立最小 Markdown 文件即可，不要求先跑其它模块。

## 开跑前

使用台账标定时，v0.2.0记录按 `application_key` 合并同岗位接触，已确认首次申请才进入申请分母；跟进原话通过同键关联，不投/待核实/预占单列。旧记录无键或事件类型时保留原材料、说明归并局限，不猜历史成功数，也不阻断当前JD定级。运行库投影与原始台账并存时不能重复统计同一申请。

### 要什么

| | 要什么 | 缺了会退化成什么 |
|---|---|---|
| **必需** | **JD 原文** | 只有名称/链接/转述时先按下文主动补齐流程检索原文；未取得前可整理探索线索，但不把转述当JD原句，不伪造抽取表 |
| **必需** | **你的必要条件与硬性排除条件（按实际有的提供，不凑数量）** | 缺了**不猜、也不用默认值**。会先要你写这两组，因为没有它们，所有岗位都会被判成"较匹配"（见「不对劲的时候」第 1 条） |
| 可选 | [`positioning.md`](../../00-workspace/templates/positioning.md)（定位文件，空模板在 `00-workspace`） | 没有就从上一行的口述里现场整理一份最小版本，并存回去，下次直接用 |
| 可选 | 加分条件 | 没有就只判"是否满足必要条件 + 是否触排除条件"，**级别会收敛成三级**（强匹配与较匹配分不开），并明确告诉你少了哪一层。⭐ **写了加分条件、但它的上沿还没校准时也走这一档**——首轮的每个人都在这个状态里，见第二步「首轮」一节 |
| 可选 | ⭐ `branches.md`（活跃探路分支） | **一律按主线阈值判，探路岗位会被判成"不投"。** 这是缺省的保守行为，不是 bug。但必须明说一句："我没看到任何活跃探路分支——**你如果在试新方向，需要手工告诉我**" |
| 可选 | `ledger.md`（台账，已跑过的岗位与对方回应） | 没有就跳过第三步的阈值反推，阈值停在 `纯占位待校准`，并在每次定级结论末尾标注这一点 |

### ⭐ 最小可接受形态

**一份JD或可检索的公司/岗位线索，加上你已有的目标条件。** 原文和条件缺失时先补齐；只得到线索不等于已完成定级。

散着贴进来就行 —— 微信里复制的一段、招聘页面上整片选中的文字、你自己口述的三句"我不接受什么"，**都算**。整理成结构化表是 agent 的工作，不是你的。

#### ⭐ 补齐原文和条件后，你会拿到 X，但拿不到 Y

| ✅ X：你会拿到 | ❌ Y：你拿不到 |
|---|---|
| 一张结构化抽取表 + 一个级别 + **能追回到条件编号的理由** | **分开的「强匹配」与「较匹配」**——没有加分条件表（或有表但上沿还没校准），级别收敛成三级 |
| 一份「原文未提及」清单，逐条升级成面试必问项 | **任何一个标着 `实践校准` 的阈值**。全部停在 `纯占位待校准`——标定要的是你**已经做过**的判断，一份新 JD 里没有 |

Report a problem