Subagent Pipeline Setup: Separating Listing Ingestion from Resume Rewriting
Ads
Asking an AI model to read a job posting and rewrite your resume in one prompt often produces fabricated experience. Setting up a subagent pipeline setup prevents this problem.
Splitting the work between one prompt that analyzes the job posting and a second prompt that audits your actual experience keeps your tailored applications grounded in real work.
Why Single-Prompt Generation Causes Hallucinations
Language models focus on matching text patterns. When given a job listing that demands specific software skills, a prompt like "tailor my resume to this job" leads the model to insert those tools directly into your past jobs.
Dividing the process into two stages stops the model from inventing qualifications:
Architecture: Stage 1 Ingestion and Stage 2 Synthesis
This process requires no code and works in any chat window:
Copy-Paste Prompts to Run This Tonight
You can run this two-stage pipeline in separate prompts or chat threads.
Step 1: The Ingestion Agent Prompt
You are a Job Requirement Auditor. Extract the following from this job description: 1) Hard technical requirements (tools, languages, frameworks). 2) Operational scale requirements (budgets, team sizes, users). 3) Soft governance expectations. Output this strictly as a Markdown bullet list. Do not write a cover letter. Do not reference any candidate. Here is the posting: [PASTE JOB DESCRIPTION]
Step 2: The Alignment and Drafting Agent Prompt
You are an Executive Resume Drafter. Below is a Target Requirement Checklist followed by my Verified Work History. Rewrite my experience bullet points to emphasize where my actual background directly matches the Target Checklist. Rules: Never invent a tool, company, metric, or responsibility not present in my Verified Work History. If a Target Requirement is missing from my history, output a note under a 'Gaps Identified' header instead of hallucinating a match. Target Checklist: [PASTE OUTPUT FROM STEP 1]. Verified Work History: [PASTE RAW RESUME].
Frequently Asked Questions
Do I need Python or an API key to use this pipeline?
No. You can run the prompts in standard chat interfaces by pasting the output of Step 1 into Step 2.
Can I feed multiple job descriptions into Agent 1 at once?
No. Run one job posting at a time. Combining multiple postings muddies the requirements and produces vague targets.
What should I do if the Drafter output highlights critical gaps?
Address missing requirements in a cover letter or interview by discussing related experience, rather than adding unverified tools to your resume.