A Three-Subagent Local Pipeline for Sourcing, Tailoring, and Interview Prep

Ads

Job seekers often write one large prompt: "Read this job description, rewrite my entire resume, write a cover letter, and give me interview questions." This broad approach fails. Large language models struggle with crowded context windows, producing generic resumes, missed requirements, and fabricated experience.

A modular workflow fixes this. By isolating the work into three separate prompts—or three chat tabs—you keep control over each stage: requirements triage, profile alignment, and interview rehearsal.

The architectural flaw of single-prompt job workflows

Language models produce better results when given a single task with explicit constraints. When one prompt demands parsing, creative writing, fact-checking, and roleplay simultaneously, the model flattens details.

Splitting the work into three steps prevents context bleed. Your resume details remain tied to your actual work, the job requirements are parsed without promotional copy, and interview practice focuses on the employer's stated operational problems.

Agent 1: The Job Listing Triager

Open a dedicated chat window. This subagent analyzes job postings to extract hard requirements and tech stack requirements while ignoring company marketing copy.

You are an executive talent sourcer. Analyze the job posting text I provide.

Output ONLY:
1. Core Hard Requirements: Must-have skills, tools, and years of experience explicitly required.
2. Nice-to-Have Skills: Secondary or preferred qualifications.
3. Knockout Criteria: Probable disqualifiers (location, clearances, required credentials).
4. Primary Operational Problem: In 2 sentences, what business problem is this hire expected to solve?

Do not rewrite my resume yet. Await the job posting text.

Agent 2: The Alignment and Bullet Tailorer

In a second window, set up an alignment editor. Give this agent the output from Agent 1 along with your master list of verified work accomplishments. Instruct it to match your documented experience to the job's core requirements without adding missing tools.

You are a technical resume writer. I will provide two inputs:
1. The core requirements from a target job listing.
2. My verified master career accomplishment database.

Your task:
- Match my existing, documented achievements to the job's core requirements.
- Output 3 tailored bullet points for my most recent role that highlight overlapping experience.
- Explicitly list any critical job requirements where I have NO documented match, labeling them 'GAPS'.
- NEVER invent tools, metrics, or responsibilities to cover gaps.

Agent 3: The Technical and Behavioral Drillmaster

In a third window, run an interview coach. Paste the role's primary problem (from Agent 1) and your tailored bullets (from Agent 2). This agent tests your ability to explain your background directly.

You are a hiring manager interviewing me for the role discussed. 

Review my tailored background bullets and the primary operational problem this role solves. 
Ask me ONE challenging, realistic interview question at a time focusing on architectural decisions, trade-offs, or leadership challenges.
Wait for my response. After I answer, grade my clarity, point out any unverified fluff, and ask the next question.

Executing the pipeline cleanly

Run these three prompts in sequence for each job application. Using separate browser tabs lets you analyze a new job posting in Agent 1 without clearing your career history in Agent 2 or resetting the interview history in Agent 3.

Frequently Asked Questions

Do I need specialized agentic software to run this?

No. You can run this pipeline in any standard LLM chat interface by keeping three separate tabs open for Agent 1, Agent 2, and Agent 3.

What if Agent 2 identifies significant skill gaps?

That is the goal of the setup. Identifying gaps before an interview lets you prepare honest explanations of how you learn new tools instead of getting caught unprepared.

Can I automate the handoff between these agents?

Yes, through API scripts, workflow tools, or local model runners. Pasting manually between tabs, however, provides an easy way to verify facts before the next step runs.