claude-cowork-job-application

KO | EN

gitlite — search

#javascript #python #golang #go #rust #docker #nodejs #java #windows #awesome #kubernetes #security #mcp #open-source

claude-cowork-job-application

(★ 12)

No description available.

claude-cowork-job-application 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# claude-cowork-job-application
## Job Application Agent

A practical prompt for finding relevant jobs, tailoring a CV to each role, writing cover letters and completing applications without letting the AI invent a better-qualified version of you.

I built this after helping my sister switch roles.

We were going through the usual process: find a job, read the description, edit the CV, rewrite the summary, prepare another cover letter and repeat the whole thing for every fucking application.

So I decided to automate it.

I gave the prompt her CV, ran it through Claude Cowork and used it to search for suitable roles, assess her fit and prepare applications. It helped her apply for roughly 38 jobs, with a tailored CV for each role and relevant cover letters where required.

The first version worked, but it also exposed the obvious problem with AI job applications: left unsupervised, the model will sometimes exaggerate experience, make convenient assumptions or confidently fill gaps it has no business filling.

This repository is the result of tightening that workflow.

## What it does

The agent can:

* Analyse a candidate’s CV
* Identify exact-match and adjacent roles
* Ask for missing information before it becomes an application blocker
* Search for live vacancies
* Assess each job against the candidate’s actual experience
* Reject roles that are a poor fit
* Tailor the CV for each approved application
* Draft role-specific cover letters
* Prepare answers to screening questions
* Complete application forms where browser access is available
* Stop before sensitive questions, declarations or unsupported claims
* Maintain an application tracker

The point is not to spray the same CV at hundreds of vacancies.

The point is to remove repetitive work while keeping the applications accurate.

## The main idea

Most AI job-search workflows start with:

> Here is my CV. Find jobs and apply for me.

That instruction is far too loose.

A reliable application agent needs more than a CV. It needs:

* A candidate profile
* A bank of confirmed evidence
* Search preferences
* Salary and location constraints
* Work-authorisation information
* Permission levels
* Rules for handling missing information
* Clear limits on what it can claim
* A record of previous applications

Without those controls, the AI may produce polished applications that the candidate cannot defend in an interview.

## Core safeguards

The prompt is designed around a few non-negotiable rules.

### It cannot invent experience

Every meaningful claim must come from the CV, portfolio, candidate profile or an answer confirmed by the candidate.

The agent must preserve the difference between:

* Supporting a project
* Contributing to a project
* Coordinating a project
* Managing a project
* Leading a project

Those are not interchangeable.

### It builds an evidence bank

Instead of collecting vague skills, the agent records the evidence behind them.

For each project or achievement, it looks for:

* The problem
* The candidate’s responsibility
* The actions taken
* The methods and tools used
* The people involved
* The scale
* The result
* The wider effect
* The source of the claim

Evidence is labelled as confirmed, partially confirmed, inferred or unsupported.

Unsupported evidence cannot be used as fact.

### It asks questions before they become blockers

A CV rarely contains everything an application form needs.

The prompt asks about likely blockers early, including:

* Salary expectations
* Notice period
* Start date
* Right to work
* Sponsorship requirements
* Preferred locations
* Remote or hybrid preferences
* Contract types
* Willingness to relocate
* Reasons for leaving previous roles

The questions are grouped together so the agent does not interrupt the user every few minutes.

### It can say “do not apply”

The agent assesses each vacancy against the candidate’s actual evidence.

It separates mandatory requirements from employer wish lists and checks:

* Core skills
* Relevant experience
* Seniority
* Demonstrated outcomes
* Industry knowledge
* Location
* Salary
* Work authorisation
* Mandatory qualifications

A failed mandatory requirement overrides the match score.

The agent is allowed to conclude that an application would be a waste of time.

### It uses permission levels

“Apply for jobs for me” can mean several things.

The prompt includes four modes:

1. **Research only**
   Find, assess and shortlist roles.

2. **Prepare applications**
   Tailor documents and draft answers without completing the form.

3. **Assisted application**
   Complete the form and upload documents, then stop before submission.

4. **Authorised submission**
   Submit suitable applications while still pausing for declarations, sensitive questions, assessments and identity checks.

The default mode is **Prepare applications**.

## What the agent must stop for

Even in authorised submission mode, the agent must pause for:

* Legal declarations
* Salary or contractual commitments
* Equality-monitoring questions
* Health or disability information
* Criminal-record questions
* Security-clearance questions
* Identity verification
* CAPTCHA
* Multi-factor authentication
* Recorded interviews
* Assessments or tests
* Questions that cannot be answered from confirmed information

Automation should remove repetitive work. It should not make personal or legal decisions on someone’s behalf.

## How to use it

### 1. Copy the prompt

Open the main prompt file:

```text
/prompts/job-application-agent.md
```

Copy the full contents into an AI assistant with access to the tools you want it to use.

The workflow is most useful with an agent that can:

* Read uploaded files
* Browse the web
* Create DOCX and PDF files
* Interact with webpages
* Maintain structured data

It can still be used without browser automation for job research, CV tailoring and application preparation.

### 2. Upload your CV

Provide the agent with your current CV.

You can also include:

* Portfolio links
* LinkedIn profile
* Previous cover letters
* Work samples
* Certifications
* A master career-history document
* Existing application answers

### 3. Complete the intake

The agent will first read the CV, extract what it already knows and ask for missing information.

Do not skip this stage.

The quality of later applications depends on the quality of the candidate profile and evidence bank.

### 4. Review the suggested role categories

The agent will identify:

* Exact-match roles
* Strong adjacent roles
* Developmental roles
* Aspirational roles
* Roles to avoid

Review these before starting a broad search.

A job title can mean very different things across industries and employers.

### 5. Choose an application mode

Choose how much authority the agent has.

For a first run, use:

```text
Prepare applications
```

Move to assisted application after you trust the profile, evidence bank and document quality.

### 6. Review before submission

Check every application for:

* Unsupported claims
* Incorrect dates
* Inflated responsibilities
* Wrong company names
* Salary errors
* Work-authorisation errors
* Generic cover-letter language
* Incorrect document uploads
* Questions the agent answered by assumption

Do not outsource judgement just because the workflow is automated.

## Suggested repository structure

```text
job-application-agent/
│
├── README.md
├── LICENSE
│
├── prompts/
│   ├── job-application-agent.md
│   ├── candidate-intake.md
│   ├── cv-tailoring-rules.md
│   └── cover-letter-rules.md
│
├── templates/
│   ├── candidate-profile.md
│   ├── evidence-bank.md
│   ├── job-assessment.md
│   └── application-tracker.csv
│
├── examples/
│   ├── example-candidate-profile.md
│   ├── example-job-assessment.md
│   └── example-workflow.md
│
└── docs/
    ├── safety-and-permissions.md
    └── browser-automation-notes.md
```

## Recommended workflow

```text
CV
  ↓
Candidate profile
  ↓
Evidence bank
  ↓
Role categories
  ↓
Search strategy
  ↓
Live job search
  ↓
Job assessment
  ↓
Candidate approval
  ↓
Tailored CV and cover letter
  ↓
Application form
  ↓
Final review
  ↓
Submission
  ↓
Application tracker
```

## Why the prompt is long

Because job applications are not a single writing task.

They involve:

* Personal information
* Employment history
* Legal eligibility
* Salary decisions
* Career positioning
* Evidence
* Document generation
* Browser interaction
* Consent
* Record keeping

A short prompt can produce a decent cover letter.

It is less likely to produce a reliable application process.

Use the sections you need, remove the ones you do not and adapt the workflow to your own situation.

## Known limitations

### Job sites are inconsistent

Many employer portals require candidates to create accounts, verify email addresses and manually re-enter information already contained in the CV.

Some sites also have poor parsing, broken uploads, expired sessions and multi-page forms that cannot be completed reliably without supervision.

LinkedIn, Indeed and similar platforms are generally easier to work with than individual applicant-tracking systems.

### Browser automation is not universal

Some AI tools cannot directly operate a browser.

Others may be blocked by:

* CAPTCHA
* Login requirements
* Verification codes
* Anti-bot controls
* Unsupported file uploads
* Session timeouts

When direct application is unavailable, the agent should prepare the full application pack and provide the verified application link.

### A match score is still a judgement

The scoring framework helps create consistency, but it is not a prediction of whether the candidate will receive an interview.

Recruitment decisions depend on the applicant pool, internal candidates, hiring-manager preferences and criteria that may not appear in the job description.

### Tailoring cannot replace missing experience

The prompt can surface relevant evidence and improve how it is presented.

It cannot turn an unqualified candidate into a qualified one.

## Privacy

Job applications contain a large amount of personal information.

Before using any AI tool, check:

* How uploaded documents are stored
* Whether conversations are used for training
* Whether browser activity is retained
* What integrations have access to your data
* Whether application documents contain unnecessary personal information

Avoid uploading:

* Passwords
* Identity documents
* Bank details
* National Insurance numbers
* Immigration documents
* Sensitive medical information

Provide sensitive information only where it is genuinely required and through the employer’s official application system.

## Contributing

Pull requests are welcome.

Useful contributions include:

* Better intake questions
* Improved scoring logic
* Additional job-board workflows
* Safer browser-automation rules
* Application tracker formats
* Support for different countries
* Accessibility improvements
* Examples from real job searches
* Better handling of career changers
* Better handling of contractors and freelancers

Do not submit examples containing real candidate information unless you have permission to publish it.

## Disclaimer

This repository is a workflow template, not a recruitment service, legal adviser or guarantee of employment.

Review every application before submission.

You remain responsible for the accuracy of the information supplied to employers.

## Licence

Choose a licence based on how you want the prompt to be reused.

For open reuse and modification, MIT is the simplest option.

## Credits

Built after helping my sister automate the most repetitive parts of her job search.

The AI helped us move faster.

The rules stopped it from getting creative with her career.

claude-cowork-job-application

claude-cowork-job-application 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

claude-cowork-job-application

claude-cowork-job-application 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README