convez.ai
Skill
agrawalankit90

job-search-project

agrawalankit90/job-search-project

Ver no GitHub
InicialPluginScore · de 0 a 100

AI-powered job search system built with Claude Cowork — strategy, tailored applications, and job tracking using CLAUDE.md context files and custom skills.

Instalação

$npx convez-ai agrawalankit90/job-search-project

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
2
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

AI-Powered Job Search System — Built with Claude Cowork

A personal job search assistant built entirely using Claude Cowork — no code required. Just a folder, markdown files, a custom AI skill, and the right prompts.

This project turns Claude into a persistent, context-aware job search partner that knows who you are, what you're targeting, and can generate tailored resumes, cover letters, outreach messages, company research, 30-day action plans, and a live job tracker — all from your own files.


Table of Contents

  1. Project Overview
  2. Architecture & How It Works
  3. Folder Structure
  4. Prerequisites
  5. Step-by-Step Tutorial
  6. How to Run It
  7. Prompt Reference (All Prompts in One Place)

1. Project Overview

Job searching is repetitive, manual, and mentally draining. This project automates the high-effort parts using Claude as an AI assistant that has full context about who you are and what you want.

What it does:

  • Maintains a persistent "memory" of your background, goals, and constraints across every Claude session — so you never have to re-explain yourself
  • Generates a personalized, data-driven job search strategy using a custom AI skill you build once and reuse forever
  • Creates company-specific packages (tailored resume, cover letter, outreach messages, company research, 30-day action plan) with a single prompt
  • Tracks all job opportunities in a structured CSV, prioritized by match percentage, location, and posting date
  • Walks you through applying to jobs — opening each link, matching the JD to your resume, getting your confirmation, and updating the tracker

Who this is for:

Anyone job searching who wants to use AI as a real productivity multiplier — not just to polish a cover letter, but to run the entire operation. Especially useful for senior professionals where every application needs to be precise and tailored.


2. Architecture & How It Works

The system is built on three concepts native to Claude Cowork: CLAUDE.md context files, custom skills, and file referencing.

Concept 1: CLAUDE.md as Persistent Memory

Claude reads any file named CLAUDE.md in your connected folder automatically at the start of every session. This is your persistent memory layer — Claude knows who you are before you type a single word.

This project uses two CLAUDE.md files at different levels:

FileLocationPurpose
Global CLAUDE.mdjob-search-project/CLAUDE.mdWho you are — background, skills, communication style, what you'll ask for
Project CLAUDE.mdjob-search-project/job-search/CLAUDE.mdJob-search-specific — target roles, target companies, constraints, working instructions

The global one sets the identity layer. The project one sets the operational layer. Together, they replace the "let me give you some context" preamble you'd otherwise type every single session.

Concept 2: Custom Skills

A skill is a reusable Claude behavior you define once in a SKILL.md file. It contains the goal, trigger conditions, step-by-step instructions for Claude to follow, and output requirements. Once saved, you activate it in any session by invoking its name.

This project includes one skill:

SkillLocationWhat It Does
job-strategy-plannerjob-strategy-planner/SKILL.mdCollects your resume + LinkedIn profile, asks follow-up questions, and generates a complete personalized job search strategy using your own template

Concept 3: File Referencing with @

In Claude Cowork, you can reference any file in your connected folder using @filename.md. Claude reads the file inline as part of your prompt. This is how you pass your resume, strategy, or job list into any instruction without copy-pasting.

For example: @my-resume.md makes Claude read your resume before acting on whatever you ask next.

How a Session Works

Session Start
    │
    ▼
Claude reads CLAUDE.md (global) + CLAUDE.md (job-search/)
    │
    ▼
You type a prompt with @file references
    │
    ├── Resume tailoring?     → Claude uses @my-resume.md + company JD
    ├── Company package?      → Plan mode builds 6 files in sequence
    ├── Job search?           → Claude searches, ranks, appends to recommended_jobs.csv
    └── Apply?                → Claude opens each link, matches JD, waits for your go/no-go

Key Design Decisions

Why markdown files, not a database? Markdown is readable, versionable (Git-friendly), and directly referenceable with @. No setup required.

Why a template for the strategy? A template enforces consistent structure across every strategy generated — whether it's for you or for someone else rebuilding this. It also makes the skill deterministic: Claude fills in real data, it doesn't hallucinate a format.

Why two CLAUDE.md files? The global one (job-search-project/) is identity — it doesn't change. The project one (job-search/) is operational — it changes as your search evolves. Separating them keeps both clean and focused.


3. Folder Structure

job-search-project/                    ← Global folder (connected to Claude Cowork)
│
├── CLAUDE.md                          ← Global identity: who you are, how you work
├── README.md                          ← This file
│
├── job-strategy-planner/
│   └── SKILL.md                       ← Custom Claude skill for job strategy generation
│
└── job-search/                        ← Working folder for active job search
    ├── CLAUDE.md                      ← Job-search context: targets, constraints, working rules
    ├── my-resume.md                   ← Resume converted from PDF to markdown
    ├── job-strategy-plan-template.md  ← Manually written strategy template (input to skill)
    ├── job-strategy-ankit-2026.md     ← Generated strategy output (produced by skill)
    ├── recommended_jobs.csv           ← Live job tracker: ranked, prioritized, with status
    │
    └── Agoda/                         ← Company-specific package (one folder per company)
        ├── agoda-company-research.md  ← Company overview, PM culture, product bets
        ├── agoda-strategy.md          ← Application strategy for this specific role
        ├── agoda-resume-ankit.md      ← Resume tailored to Agoda's JD
        ├── agoda-cover-letter.md      ← Cover letter (~215 words)
        ├── agoda-outreach.md          ← LinkedIn DM and email variants (A/B/C)
        └── agoda-30-day-plan.md       ← Day-by-day action plan: apply → offer

Pattern to follow: Every company you target gets its own subfolder inside job-search/. Each subfolder gets the same 6-file package generated by one Plan Mode prompt.


4. Prerequisites

  • Claude Desktop App with Cowork mode enabled
  • A resume in PDF format
  • A LinkedIn profile (URL is enough — Claude can reference it)
  • No coding knowledge required. No installations. No APIs.

5. Step-by-Step Tutorial

This section walks you through exactly how this project was built — step by step, including the exact prompts used at each stage.


Step 1 — Set Up Your Workspace

Create two folders on your computer:

job-search-project/       ← Connect this to Claude Cowork as your global folder
└── job-search/           ← This is your working folder for active search

In Claude Cowork, select job-search-project as your connected folder. Claude will automatically read any CLAUDE.md file it finds here at the start of every session.


Step 2 — Convert Your Resume to Markdown

Why: Claude works best with text-based files it can read and reference inline. Converting your PDF resume to markdown makes it referenceable with @my-resume.md in any prompt.

How:

  1. Open Claude Cowork
  2. Upload your resume.pdf using the attachment button
  3. Type this prompt:
Create a resume.md file for the resume.pdf file provided.

Claude will extract all content from your PDF and save a clean my-resume.md file in your connected folder.

Output: job-search/my-resume.md


Step 3 — Create the Global CLAUDE.md (Who You Are)

Why: This is the identity layer — the file Claude reads first in every session. It tells Claude your background, expertise, goals, and communication preferences so you never have to explain yourself again.

How:

With my-resume.md already in your folder, type this prompt:

@my-resume.md Use my resume above to write my CLAUDE.md — the file that tells Claude who I am, 
how I work, and what I'm trying to accomplish, so every future session starts with full context about me. Include:

* My background and current role/stage
* My core skills and domain expertise
* My career goals and what I'm optimizing for
* How I like to work and communicate (concise, structured, no fluff)
* What kind of help I'll typically be asking for

Write this as a tight, scannable CLAUDE.md — bullet points, no filler. Claude should be able to read 
this and immediately know who I am and how to be most useful to me. After writing it, save the result 
to job-search-project.

Output: job-search-project/CLAUDE.md

What it produces: A structured profile with sections for who you are, your domain expertise, career goals, communication preferences, and the types of tasks you'll bring to Claude.

Tip: After Claude generates this file, read it and edit anything that's off. This file persists forever — the more accurate it is, the more useful every future session becomes.


Step 4 — Create the Project CLAUDE.md (Job Search Context)

Why: The global CLAUDE.md handles identity. This one handles operations — it tells Claude what you're targeting right now: specific roles, companies, timeline, location, comp floor. It also contains working instructions so Claude never gives you generic advice.

How:

This prompt triggers an interactive flow — Claude asks you four questions before writing the file:

@my-resume.md Using my resume above, write the local CLAUDE.md for my current job search project. 
Before writing anything, use the askUserQuestion tool to ask:

1. target timeline — when do I need an offer by?
2. location preference — remote, specific cities, or open to relocation?
3. compensation floor — minimum base salary I'll consider
4. current search stage — just starting, actively applying, or already interviewing?

Then write the CLAUDE.md using both my resume and my answers. Include:

* My target roles (be specific — titles I should be applying for given my background)
* Target companies (if I haven't specified, suggest 5-8 that match my profile)
* My constraints: timeline, location, compensation floor, search stage
* Working instructions for this project: when I ask for job strategy, always reference my actual 
  experience. Don't give generic advice.

Save the result to job-search in this folder.

Claude will ask you:

  • When do you need an offer?
  • Where are you open to working?
  • What's your minimum acceptable salary?
  • What's your current application stage?

Answer these directly in the chat. Claude uses your answers + your resume to write a job-search-specific CLAUDE.md with your target role titles, a company list tiered by fit, your constraints as a table, and working instructions that force Claude to ground every recommendation in your actual experience.

Output: job-search/CLAUDE.md


Step 5 — Write Your Job Strategy Template

Why: The strategy template defines the structure of every job search strategy generated by your custom skill. Writing it once gives you a repeatable, consistent output every time the skill runs — for you or for anyone rebuilding this system.

How: Write this file manually. It is not generated by Claude — it's your own document defining what a good strategy looks like.

The template should include sections for:

  • Strategy Statement — one-sentence goal (role, level, timeline, compensation target)
  • Step 1: Diagnosis — current situation, challenges, opportunities, market/competitive analysis
  • Step 2: Guiding Policy — overall approach, core principles (non-negotiables), long-term goal, SMART short-term objectives
  • Step 3: Action Plan & Roadmap — key initiatives, resource allocation, support system, milestones table, weekly KPIs
  • Skill Gap Analysis — a table of gaps, specific actions to close each

Use placeholder text (like [User's Name] or [fill in]) wherever user-specific data will be inserted by Claude. Save the file as:

Output: job-search/job-strategy-plan-template.md

See the actual template in this repo at job-search/job-strategy-plan-template.md for a complete example you can copy directly.


Step 6 — Build the Job Strategy Planner Skill

Why: A skill is a reusable Claude behavior. Instead of writing a long prompt every time you want a job strategy, you invoke the skill once and Claude knows exactly what to collect and how to format the output.

How:

With job-strategy-plan-template.md in your folder, use the skill-creator tool in Claude Cowork with this prompt:

@job-strategy-plan-template.md Create a skill named job-strategy-planner:

Goal
This skill helps users generate personalized job search strategies using the template attached to you.

Behavior
Trigger this skill whenever a user asks about:
* Job strategy
* Career planning
* Job switch
* Job preparation
* Breaking into a role/company

Instructions
* Use the askUserQuestion tool to gather required information from the user.
* You MUST collect the following before generating any strategy:
   * Resume (text or file)
   * LinkedIn profile
* Do NOT generate any strategy until both Resume and LinkedIn profile are provided.
* After collecting the required inputs, continue asking follow-up questions using the askUserQuestion 
  tool to improve context, such as:
   * Target role(s)
   * Target company (if any)
   * Experience level
   * Constraints (time, location, timeline, etc.)
* Keep asking questions until you have sufficient context.

Output Requirements
The final output MUST strictly follow the structure and format of the provided template 
(@job-strategy-plan-template.md). Do NOT deviate from the template format. Replace template 
placeholders with user-specific data only. After generating the strategy, save the final output 
to a file (e.g., job-strategy.md).

Rules
Do NOT perform any evaluation or test case — just create a skill.
Do NOT generate generic responses.
Always prioritize better context before answering.
Think like a career coach + recruiter.

Output: job-strategy-planner/SKILL.md

This creates a skill Claude can invoke in any future session. The skill enforces: collect resume first, collect LinkedIn second, ask follow-up questions, then and only then generate a strategy that strictly follows your template.


Step 7 — Run the Skill to Generate Your Strategy

Why: This is where the skill pays off. You kick it off with a rough strategy statement, Claude collects all required inputs interactively, then produces a complete, personalized strategy saved to a file.

How:

In a new Claude session, type:

I will transition into a [role] at [level] by building real AI products in [domain], creating strong 
public proof of work, and consistently applying with targeted signals — aiming to secure the role 
in [X months] and reach [$X compensation] within [timeframe].

Fill in your actual blanks. Claude recognizes this as a job strategy trigger, invokes the job-strategy-planner skill, and begins its collection flow:

  1. Asks for your resume (reference @my-resume.md or paste it)
  2. Asks for your LinkedIn profile URL
  3. Asks follow-up questions (target roles, companies, constraints, differentiators)
  4. Generates the strategy using your template's exact structure — filled with your real data
  5. Saves the file automatically

Output: job-search/job-strategy-ankit-2026.md (or whatever name you specify)

What the output contains: A strategy statement, situation diagnosis, challenges and opportunities, tiered target company list (20–40 companies), guiding policy, SMART objectives table, 8-week implementation roadmap, weekly KPIs to track, and a skill gap analysis with specific actions for each gap.

Tip: The strategy file becomes a key input to everything else — the job finder, company packages, and cover letters all reference it. The more specific your answers during the skill's question flow, the sharper the output.


Step 8 — Build a Company-Specific Package (Plan Mode)

Why: Each serious application deserves a tailored package: company research, a repositioned resume, a tight cover letter, outreach messages for LinkedIn and email, and a day-by-day action plan from application to offer. Doing this manually would take hours. Plan Mode does it in one approved prompt.

How:

Plan Mode in Claude Cowork lets Claude lay out its full plan before doing any work. You review it, approve it, and then Claude executes — creating all files in the right order.

Use this prompt:

@my-resume.md Plan mode: I want you to build me a complete job search package for a Senior Product 
Manager role at [Company Name]. Before doing anything, lay out your plan — what you'll create, in 
what order, and what files you'll save. Wait for my approval before starting.

Claude will respond with a plan listing all 6 files it intends to create and the order of operations. Once you approve (type "yes" or "looks good"), Claude executes the plan and saves all files to a company-specific folder.

Output for each company (e.g., job-search/Agoda/):

FileContents
agoda-company-research.mdCompany overview, recent product bets, PM culture signals, how your background maps to their priorities, talking points for interviews, red flags
agoda-strategy.mdPositioning angle, application channel priority (referral → direct → cold), gap bridge plan, competitive landscape, non-negotiables
agoda-resume-ankit.mdFull resume with summary rewritten to lead with the most relevant proof points for this specific role
agoda-cover-letter.md~200-word cover letter with a specific hook, your strongest relevant proof point, and a clear bridge to their current product bets
agoda-outreach.md3 outreach message variants: LinkedIn DM to recruiter, cold email, LinkedIn DM via referral — each under 80 words
agoda-30-day-plan.mdWeek-by-week action table from Day 1 (apply + activate referrals) to Day 30 (offer decision), with daily actions, time estimates, and impact ratings

To build packages for more companies: Repeat this step with a different company name. Each company gets its own folder.


Step 9 — Find & Track Jobs

Why: Instead of manually scrolling job boards, you can have Claude search for open roles, match them against your resume and strategy, rank them, and maintain a structured tracker — all in one prompt.

How:

You have access to the following inputs:

1. My resume (@my-resume.md) — skills, experience, roles, preferences
2. My job search strategy (@job-strategy-ankit-2026.md) — target roles, locations, constraints, priorities
3. A @recommended_jobs.csv file containing jobs with status applied. Ignore if not present.

Your task:
* Analyze my resume to understand my skills, experience level, and role fit
* Use my job search strategy to define what "relevant jobs" mean for me
* Read the CSV file and identify companies, roles, and job types I have already applied to
* Avoid suggesting duplicate roles or the same company + role combinations

Now do the following:
1. Search for new and relevant job openings that match my resume and strategy
2. Prioritize jobs posted in the last 24–48 hours
3. Rank jobs by relevance (best fit first)
4. For each job, extract:
   * Company Name
   * Job Title
   * Location
   * Job Type (remote / hybrid / onsite)
   * Experience Level
   * Job Posting Date
   * Job URL
   * Short reason why this role is a good match for me
   * Short description on what it takes to be a best match
   * Priority based on: Match % (>70% required) → Location → Posting recency

Output:
* Create a CSV file named recommended_jobs.csv. If already present, append new jobs to the end.
* Store all selected job listings in this CSV.

Output: job-search/recommended_jobs.csv

CSV columns: Rank, Priority (P1/P2/P3/P4), Company Name, Job Title, Location, Job Type, Experience Level, Job Posting Date, Match %, Job URL, Why Good Match, What Is Missing / Gap, Priority Rationale, Status (applied / blank)

Priority logic:

  • P1 – Apply Now: Match >80% + strong location + recent posting
  • P2 – Apply This Week: Match 70–80% OR slight level/location mismatch
  • P3 – Apply If Bandwidth: Match 60–70% OR strategic but not urgent
  • P4 – Deprioritize: Match <70% OR hard constraint mismatch

Run this prompt regularly — every few days — to add fresh postings. Claude reads the existing CSV, skips companies you've already applied to, and appends new rows.


Step 10 — Apply for Jobs

Why: The tracker is only useful if you act on it. This prompt turns the CSV into an active application workflow — Claude opens each job link, does a fresh JD-to-resume match, shows you the quality assessment, waits for your go/no-go, and updates the tracker as you go.

How:

Upload your resume.pdf (for form submissions) alongside this prompt:

Use the file @recommended_jobs.csv and my resume @resume.pdf. Pick the combination of:
1. Non-applied jobs
2. P1 priority

* Open each job link (filtered based on the above combination) using the browser tool, and if the 
  Job Description is available/accessible:
  * Do a thorough match of the job against the resume
  * Give a brief description on the quality of match and what is missing or non-negotiable
  * Get a confirmation from me if I want to continue to apply
  * If my answer is yes, apply to the relevant role
  * If my answer is no, use the info to update the respective columns (Priority, Match %, Why Good 
    Match, What Is Missing) for that job in @recommended_jobs.csv. Then move to the next job link.

* If the JD is not available, analyse my resume and match it with the job data already in the CSV.

How to apply:
* Guide me through submitting each application, or complete the steps you can do yourself
* If a job page doesn't load or isn't found, skip it and move to the next job link

Important: Before moving to the next job link, update @recommended_jobs.csv with status "applied" 
for any job where the application was submitted.

What happens in practice:

  1. Claude opens the first P1 job link in your browser
  2. Claude reads the JD and matches it against your resume — shows match %, strengths, gaps
  3. You type "yes" (apply) or "no" (skip)
  4. If yes: Claude guides you through the form or completes what it can
  5. If no: Claude updates the CSV with revised match/gap notes and moves to the next job
  6. After each application, the CSV is updated with status: applied
  7. Repeat until all P1 jobs are processed

6. How to Run It

First-Time Setup (one time only)

  1. Create the folder structure:

    job-search-project/
    └── job-search/
    
  2. Open Claude Desktop App → Cowork mode → connect job-search-project as your folder

  3. Upload your resume.pdf and run Step 2 to create my-resume.md

  4. Run Step 3 to create the global CLAUDE.md

  5. Run Step 4 to create the job-search CLAUDE.md

  6. Write job-strategy-plan-template.md manually (Step 5) — copy the template in this repo as a starting point

  7. Run Step 6 to build the job-strategy-planner skill

  8. Run Step 7 to generate your personalized strategy

Daily Usage (recurring)

TaskPrompt to useOutput
Find new job openingsStep 9 promptAppends to recommended_jobs.csv
Apply to P1 jobsStep 10 promptUpdates CSV with applied status
Build a company packageStep 8 prompt (Plan Mode)New folder with 6 files
Update your strategyRe-run Step 7 skillUpdated strategy file

Updating Your Context

When your situation changes (new constraint, updated comp floor, changed target companies), edit the relevant CLAUDE.md file directly. The changes take effect in the next session automatically — no re-prompting needed.


7. Prompt Reference (All Prompts in One Place)

A quick reference for every prompt used in this project.


Convert resume PDF to markdown

Create a resume.md file for the resume.pdf file provided.

Create global CLAUDE.md

@my-resume.md Use my resume above to write my CLAUDE.md — the file that tells Claude who I am, 
how I work, and what I'm trying to accomplish, so every future session starts with full context about me. Include:

* My background and current role/stage
* My core skills and domain expertise
* My career goals and what I'm optimizing for
* How I like to work and communicate (concise, structured, no fluff)
* What kind of help I'll typically be asking for

Write this as a tight, scannable CLAUDE.md — bullet points, no filler. Claude should be able to 
read this and immediately know who I am and how to be most useful to me.
After writing it, save the result to job-search-project.

Create project CLAUDE.md (interactive)

@my-resume.md Using my resume above, write the local CLAUDE.md for my current job search project. 
Before writing anything, use the askUserQuestion tool to ask:

1. target timeline — when do I need an offer by?
2. location preference — remote, specific cities, or open to relocation?
3. compensation floor — minimum base salary I'll consider
4. current search stage — just starting, actively applying, or already interviewing?

Then write the CLAUDE.md using both my resume and my answers. Include:

* My target roles (be specific — titles I should be applying for given my background)
* Target companies (if I haven't specified, suggest 5-8 that match my profile)
* My constraints: timeline, location, compensation floor, search stage
* Working instructions for this project: when I ask for job strategy, always reference my actual 
  experience. Don't give generic advice.

Save the result to job-search in this folder.

Create the job-strategy-planner skill

@job-strategy-plan-template.md Create a skill named job-strategy-planner:

Goal
This skill helps users generate personalized job search strategies using the template attached to you.

Behavior
Trigger this skill whenever a user asks about:
* Job strategy
* Career planning
* Job switch
* Job preparation
* Breaking into a role/company

Instructions
* Use the askUserQuestion tool to gather required information from the user.
* You MUST collect the following before generating any strategy:
   * Resume (text or file)
   * LinkedIn profile
* Do NOT generate any strategy until both Resume and LinkedIn profile are provided.
* After collecting the required inputs, continue asking follow-up questions using the askUserQuestion 
  tool to improve context, such as:
   * Target role(s)
   * Target company (if any)
   * Experience level
   * Constraints (time, location, timeline, etc.)
* Keep asking questions until you have sufficient context.

Output Requirements
The final output MUST strictly follow the structure and format of the provided template 
(@job-strategy-plan-template.md). Do NOT deviate from the template format. Replace template 
placeholders with user-specific data only. After generating the strategy, save the final output 
to a file (e.g., job-strategy.md).

Rules
Do NOT perform any evaluation or test case — just create a skill.
Do NOT generate generic responses.
Always prioritize better context before answering.
Think like a career coach + recruiter.

Run the job-strategy-planner skill

I will transition into a [role] at [level] by building real AI products in [domain], creating 
strong public proof of work, and consistently applying with targeted signals — aiming to secure 
the role in [X months] and reach [$X compensation] within [timeframe].

Build a company-specific job search package (Plan Mode)

@my-resume.md Plan mode: I want you to build me a complete job search package for a Senior Product 
Manager role at [Company Name]. Before doing anything, lay out your plan — what you'll create, in 
what order, and what files you'll save. Wait for my approval before starting.

Find and track new job openings

You have access to the following inputs:

1. My resume (@my-resume.md) — skills, experience, roles, preferences
2. My job search strategy (@job-strategy-ankit-2026.md) — target roles, locations, constraints, priorities
3. A @recommended_jobs.csv file containing jobs with status applied. Ignore if not present.

Your task:
* Analyze my resume to understand my skills, experience level, and role fit
* Use my job search strategy to define what "relevant jobs" mean for me
* Read the CSV file and identify companies, roles, and job types I have already applied to
* Avoid suggesting duplicate roles or the same company + role combinations

Now do the following:
1. Search for new and relevant job openings that match my resume and strategy
2. Prioritize jobs posted in the last 24–48 hours
3. Rank jobs by relevance (best fit first)
4. For each job, extract:
   * Company Name
   * Job Title
   * Location
   * Job Type (remote / hybrid / onsite)
   * Experience Level
   * Job Posting Date
   * Job URL
   * Short reason why this role is a good match for me
   * Short description on what it takes to be a best match
   * Priority based on: Match % (>70% required) → Location → Posting recency

Output:
* Create a CSV file named recommended_jobs.csv. If already present, append new jobs to the end.
* Store all selected job listings in this CSV.

Apply to P1 jobs from the tracker

Use the file @recommended_jobs.csv and my resume @resume.pdf. Pick the combination of:
1. Non-applied jobs
2. P1 priority

* Open each job link (filtered based on the above combination) using the browser tool, and if the 
  Job Description is available/accessible:
  * Do a thorough match of the job against the resume
  * Give a brief description on the quality of match and what is missing or non-negotiable
  * Get a confirmation from me if I want to continue to apply
  * If my answer is yes, apply to the relevant role
  * If my answer is no, use the info to update the respective columns (Priority, Match %, Why Good 
    Match, What Is Missing) for that job in @recommended_jobs.csv. Then move to the next job link.

* If the JD is not available, analyse my resume and match it with the job data already in the CSV.

How to apply:
* Guide me through submitting each application, or complete the steps you can do yourself
* If a job page doesn't load or isn't found, skip it and move to the next job link

Important: Before moving to the next job link, update @recommended_jobs.csv with status "applied" 
for any job where the application was submitted.

Built by Ankit Agrawal · June 2026 · Using Claude Cowork

Ver o original no GitHub

Bom para

Relacionados