Project 2 — Build Your AI Resume Optimizer
You will build a complete AI-powered resume optimizer using Claude — an ATS checker,
resume improver, and job-match analyzer, all driven by prompts you write yourself. This is your
first real multi-tool AI system.
What You Build
3-Tool AI System
ATS checker + Resume rewriter + Job description
matcher — all in one Claude dashboard.
What You Learn
4 Core Concepts
Structured prompting, chain prompting, context
injection, and role prompting — the backbone of AI system design.
Time to Complete
3–4 Hours
5 guided build steps. Interactive prompt
playground included. Run every prompt in Claude alongside.
Deliverables
5 Prompt Files
5 production-ready prompt files + a GitHub repo
+ your certificate of completion.
The Final System You Are Building
By the end of this module, you
will have a personal AI resume system made of 3 specialised Claude tools — each built from a
different prompting technique:
Tool 1
🎯 ATS Scanner
Scores your resume against ATS algorithms.
Finds missing keywords, flags formatting issues, predicts pass/fail rate. Technique:
Structured prompting
Tool 2
✍️ Resume Rewriter
Rewrites weak bullet points into
achievement-focused, quantified statements. Adapts tone to your target role. Technique:
Chain prompting
Tool 3
🔗 Job Matcher
Takes a job description and your resume,
finds gaps, tells you what to emphasise, generates a tailored version. Technique: Context injection + Role prompting
Your 5-Step Journey
1
Prepare your resume content
Structure your raw
resume into a clean content block — the foundation every prompt will use.
2
Build the ATS Scanner (Structured
prompting)
Write a structured
prompt that makes Claude analyse your resume like an ATS system would.
3
Build the Resume Rewriter (Chain prompting)
Build a 3-step chain
that critiques, rewrites, then scores your resume bullets — in sequence.
4
Build the Job Matcher (Context injection +
Role prompting)
Feed Claude both your
resume and a job description, inject context, and generate a tailored resume.
5
Build a Live Dashboard Inside Claude
Use Claude Artifacts to build
an interactive resume optimizer dashboard that runs live inside Claude — no files needed.
Before you start: Open claude.ai in a separate tab — you'll run every prompt there.
Keep this module open alongside it. All prompts are pre-written and copyable — your only job is to
fill in your own resume content and run them.
Concepts — The 4 Techniques You'll Master
This module introduces 4 core prompting techniques. Learn them here. You'll use each
one in the build section — one at a time, with a real tool to show for it.
Structured Prompting
Breaking your prompt into clearly labelled sections (ROLE, TASK, INPUT,
OUTPUT FORMAT, RULES). Claude reads each section as a distinct instruction layer — like
giving it a spec sheet, not a sentence.
CLICK TO DEEP DIVE →
Chain Prompting
Sending 2–4 prompts in sequence, where each one builds on the previous
output. Prompt 1 critiques. Prompt 2 rewrites using that critique. Prompt 3 scores the
rewrite. Each prompt has a single job.
CLICK TO DEEP DIVE →
Context Injection
Deliberately pasting external data (your resume text, a job description)
directly into the prompt. Claude can't see files or browse the web — you must inject
everything it needs. The more context, the better the output.
CLICK TO DEEP DIVE →
Role Prompting
Assigning Claude an expert identity before the task. "You are a Head of
Talent Acquisition at a FAANG company with 12 years experience." This activates specific
knowledge, standards, and reasoning patterns you wouldn't get from a generic prompt.
CLICK TO DEEP DIVE →
Why these 4 techniques work together
Each technique solves a
different failure mode of basic prompting:
Problem
Output is disorganised — Claude writes a
paragraph when you want a table, or gives 500 words when you want 50.
Fix
Structured prompting — tell Claude exactly what
format to use in an OUTPUT FORMAT section. Bullet points, JSON, tables, numbered list —
specify it.
Problem
Claude gives a mediocre rewrite because
it's doing critique + rewrite + scoring all in one shot and rushing.
Fix
Chain
prompting — split the task into 3 separate prompts. Let Claude fully
complete each stage before moving to the next.
Problem
Claude's output is generic — it doesn't
know YOUR resume or the specific job you're applying for.
Fix
Context
injection — paste your actual resume text and the actual job description
directly into the prompt. Don't describe them — paste them.
Problem
Feedback is too gentle, too vague, or
written by someone who doesn't know the industry.
Fix
Role
prompting — make Claude a senior recruiter at the exact company type you're
targeting. A recruiter at a FAANG thinks differently to one at a startup. Be specific.
What a Chain Prompt looks like
Here is the 3-step chain you
will build for the Resume Rewriter. See how each prompt hands off to the next:
Prompt 1
Critique bullets
Critique bullets
→
Prompt 2
Rewrite using critique
Rewrite using critique
→
Prompt
3
Score the rewrite
Score the rewrite
Key insight: Each prompt in a chain
has ONE job. When you force Claude to critique AND rewrite AND score in a single prompt, it does
all three poorly. One job per prompt = three excellent outputs.
Build It Live — 5 Steps
Follow each step in order. Run every prompt in Claude, then come back here. Each step
builds one part of your resume optimizer system.
1
Prepare Your Resume Content Block
Your raw material — every prompt depends on this
Active
Why this matters: Claude knows nothing about you. Your
name, your work history, your bullet points — none of it exists until you paste it in.
This step creates your resume content block: a structured version of your
resume that you will inject into every prompt in this module.
Fill in this template — replace every [PLACEHOLDER]
Copy the block below
into a text file. Replace every placeholder with your real information. Save it — you
will paste it into 3 different prompts.
MY RESUME CONTENT BLOCK
========================
PERSONAL INFO
Name: [Full Name]
Target Role: [e.g., "Data Analyst", "Software Engineer", "Product Manager"]
Target Industry: [e.g., "Fintech", "Pharma", "SaaS", "Consulting"]
Location: [City, Country]
Email: [your@email.com]
LinkedIn: [linkedin.com/in/yourhandle]
GitHub: [github.com/yourusername — if applicable]
SUMMARY (optional — 2 sentences)
[Your current summary section, or leave blank if you don't have one]
EXPERIENCE
Job 1:
Title: [Job Title]
Company: [Company Name]
Duration: [Month Year — Month Year]
Bullets:
- [Bullet 1 — paste exactly as it appears on your CV]
- [Bullet 2]
- [Bullet 3]
Job 2:
Title: [Job Title]
Company: [Company Name]
Duration: [Month Year — Month Year]
Bullets:
- [Bullet 1]
- [Bullet 2]
[Add more jobs if applicable]
EDUCATION
Degree: [e.g., B.Tech Computer Science / BSc Chemistry]
University: [University Name]
Year: [Graduation year]
Relevant courses: [List 3-5 relevant modules]
GPA / Grade: [If strong, include it]
SKILLS
Technical skills: [comma-separated list]
Tools / Software: [comma-separated list]
Certifications: [Any relevant certs — e.g., Google Data Analytics, AWS Cloud
Practitioner]
ACHIEVEMENTS (optional)
- [Any awards, publications, competitions, open source contributions]
RESUME FORMAT
Pages: [1 page / 2 pages]
Current file format: [PDF / Word / Google Docs]
Template style: [Minimal / Creative / Corporate / Academic]
Don't have strong bullet points yet?
Run this prompt first
to generate bullets from rough notes about what you did at a job:
I want to write professional resume bullet points for a job I held.
Here is what I actually did at this role (rough notes):
[Paste your raw notes — even fragments like "managed social media", "helped with
reports", "worked on a Python script for data cleaning"]
Role I held: [Job Title] at [Company]
Duration: [dates]
Target role I am now applying for: [Your target job]
Transform my rough notes into 4 strong, achievement-focused resume bullet points.
Rules for each bullet:
- Start with a strong action verb (Led, Built, Reduced, Improved, Automated, Delivered)
- Include a number or result where possible (%, time saved, users, revenue, etc.)
- Be specific — not "helped with" but "owned and delivered"
- Keep under 15 words per bullet
- No soft skill buzzwords (passionate, hardworking, team player)
I have filled in my Resume Content Block with real
information
I have at least 2 jobs or 2 experiences listed (internship,
project, freelance counts)
I have saved my Content Block somewhere (text file, Notion,
etc.)
2
Build the ATS Scanner
Structured prompting — your first real prompt architecture
Locked
What is ATS? Applicant Tracking Systems are software that
companies use to automatically filter resumes before a human ever reads them. 75% of
resumes are rejected by ATS before reaching a recruiter. This tool scans your resume the
same way ATS does — and tells you exactly what to fix.
Understanding Structured Prompting first
A structured prompt
has clearly labelled sections — like a spec sheet. Compare these two approaches:
❌ Unstructured prompt
Can you check my resume and tell me if it's good for ATS systems? I'm applying for
data analyst roles. Here's my resume: [big block of text]
✓ Structured prompt
[ROLE]
You are an ATS expert...
[TASK]
Analyse the resume below...
[INPUT]
[RESUME TEXT HERE]
[TARGET ROLE]
Data Analyst at a fintech company
[OUTPUT FORMAT]
1. ATS Score (0-100)
2. Missing keywords...
Why it matters: The structured version tells Claude
exactly what it is, what to do, what data to use, and how to format the answer. Each
labelled section removes ambiguity. Claude never has to guess what you want.
Your ATS Scanner Prompt — copy, fill in, run in Claude
Replace
[YOUR RESUME CONTENT BLOCK HERE]
with your content from Step 1, and [TARGET ROLE]
with your specific target.[ROLE]
You are a senior ATS specialist and recruitment consultant with 10 years of experience
screening thousands of resumes for Fortune 500 companies and fast-growing startups. You
understand exactly how ATS software parses, scores, and filters resumes before humans
see them.
[TASK]
Perform a comprehensive ATS audit on the resume below. Be direct, specific, and
actionable. Do not be encouraging for the sake of it — I need an honest score.
[MY RESUME]
[YOUR RESUME CONTENT BLOCK HERE]
[TARGET ROLE]
[e.g., "Data Analyst at a fintech startup"]
[OUTPUT FORMAT — follow this exactly]
## ATS SCORE: [X/100]
## KEYWORD ANALYSIS
Missing critical keywords (that should appear but don't):
- [keyword] — why it matters for this role
- [keyword] — why it matters for this role
[list up to 8]
Keywords found (good):
- [keyword], [keyword], [keyword]
## FORMAT ISSUES (things ATS software struggles to parse)
- [Issue 1 — e.g., "Tables detected — ATS cannot read table content"]
- [Issue 2]
[list all issues, or write "None detected" if clean]
## BULLET POINT AUDIT
Rate each bullet point from your experience section:
[Job title at Company]:
Bullet 1: "[first 5 words...]" → Score: X/10 → Issue: [what's wrong]
Bullet 2: "[first 5 words...]" → Score: X/10 → Issue: [what's wrong]
[repeat for each job]
## TOP 3 IMPROVEMENTS (ranked by impact)
1. [Most impactful fix — specific and actionable]
2. [Second most impactful fix]
3. [Third most impactful fix]
## PASS/FAIL PREDICTION
Would this resume pass an ATS filter for [TARGET ROLE]? [YES/NO/BORDERLINE]
Reasoning: [1-2 sentences]
What to do with the output
1. Read your ATS Score. Anything below 70 needs immediate attention.
2. Copy the "Missing critical keywords" list — you will add these in Step 3.
3. Note every FORMAT ISSUE — these are things you fix in your actual resume file.
4. Save this output as
2. Copy the "Missing critical keywords" list — you will add these in Step 3.
3. Note every FORMAT ISSUE — these are things you fix in your actual resume file.
4. Save this output as
ats-audit.md
— it's your first deliverable file.
I ran the ATS Scanner prompt in Claude with my resume pasted
in
I received an ATS score and a list of missing keywords
I saved the output as ats-audit.md
3
Build the Resume Rewriter
Chain prompting — 3 prompts, 3 jobs, one perfect output
Locked
The chain rule: Start a fresh Claude conversation for this
tool. Send Prompt A first. Wait for the full output. Then send Prompt B. Then Prompt C.
Do not skip steps — the quality of each output depends on the previous one.
Prompt A
Critique your bullets
Critique your bullets
→
Prompt B
Rewrite using critique
Rewrite using critique
→
Prompt
C
Score the rewrite
Score the rewrite
A
Prompt A — The Critic
Run this first
This prompt asks
Claude to be brutally honest about what is weak in your current bullets — before any
rewriting happens.
[ROLE]
You are a brutally honest senior recruiter at a top-tier consulting firm. Your job
is to identify exactly why a resume bullet point would fail to impress a hiring
manager or pass ATS screening. You do not soften criticism. You are specific, not
vague.
[TASK]
Critique every bullet point in my experience section. For each bullet, give:
1. A weakness score (1-5, where 1 = terrible, 5 = excellent)
2. The specific problem (choose one: no quantification, weak verb, vague impact, too
long, generic, missing context)
3. One-sentence explanation of exactly what is wrong
[MY EXPERIENCE BULLETS]
[PASTE YOUR EXPERIENCE BULLETS FROM YOUR CONTENT BLOCK]
[OUTPUT FORMAT]
## BULLET CRITIQUE REPORT
[Job Title at Company]
• "[First 6 words of bullet]..." → Score: X/5 | Problem: [category] | Why: [one
sentence]
• "[First 6 words...]..." → Score: X/5 | Problem: [category] | Why: [one sentence]
[repeat for all bullets]
Overall weakest bullet: "[quote it]"
Overall strongest bullet: "[quote it]"
Top 2 patterns of weakness across all bullets: [list them]
B
Prompt B — The Rewriter
Send after Prompt
A output arrives
Send this in
the same conversation after Prompt A completes. Do not paste your bullets
again — Claude already has them in context.
Now rewrite every bullet point using your critique above.
[REWRITING RULES — follow all of them]
1. Start every bullet with a strong past-tense action verb (Led, Built, Reduced,
Increased, Automated, Delivered, Designed, Launched, Implemented, Negotiated)
2. Include a quantified result where at all possible — a number, percentage, time
saved, revenue, users, accuracy rate
3. Use this formula: [Strong verb] + [what you did] + [measurable result/impact]
4. Maximum 20 words per bullet
5. No first person (no "I" or "my")
6. No clichés: passionate, hardworking, team player, results-driven, detail-oriented
7. If I haven't provided a number, invent a plausible approximate range in
[brackets] and I will fill in the real number
[TARGET ROLE]
[Your target role — e.g., "Data Analyst at a fintech company"]
[OUTPUT FORMAT]
For each bullet, show:
ORIGINAL: "[original bullet]"
REWRITTEN: "[new bullet]"
IMPROVEMENT: [one-sentence explanation of what changed and why]
Then at the end:
SUMMARY: X out of Y bullets significantly improved. Key technique used: [note the
main improvement pattern]
C
Prompt C — The Scorer
Final step in
chain
Send this after
Prompt B completes — still in the same conversation. This gives you a clear
before/after comparison score.
Score the improvement. Compare the original bullets versus your
rewrites.
[OUTPUT FORMAT]
## RESUME REWRITE SCORECARD
| Metric | Before | After |
|--------|--------|-------|
| Average bullet score (1-10) | X | X |
| Bullets with quantified results | X/Y | X/Y |
| Weak verb usage | X bullets | X bullets |
| ATS keyword density | Low/Medium/High | Low/Medium/High |
| Overall resume strength | X/100 | X/100 |
## FINAL ASSESSMENT
Biggest single improvement: [what changed most]
Remaining weakness: [what still needs work]
One more thing to add: [the single most impactful thing I could still do to this
resume]
## YOUR 3 STRONGEST NEW BULLETS
1. "[bullet]"
2. "[bullet]"
3. "[bullet]"
Save these. These are interview conversation starters.
Prompt A complete — I have a critique report for every
bullet
Prompt B complete — I have rewritten versions of every
bullet
Prompt C complete — I have a before/after scorecard saved
4
Build the Job Matcher
Context injection + role prompting — the most powerful tool in
your system
Locked
The most powerful tool of the three: This prompt takes a
real job description and your real resume and produces a tailored resume version —
specifically adapted for that one job. Most candidates submit the same resume to every
role. This is how you stop doing that.
What context injection looks like here
You will paste TWO
blocks of content into this prompt — your resume AND a job description. Claude will
analyse both simultaneously and find the gaps, matches, and opportunities:
Context 1 → Your rewritten resume (from Step
3)
Context 2 → A real job description (copy from LinkedIn, company website, etc.)
Role → A senior recruiter from that exact company type
Task → Tell me what to emphasise, what to add, what to change — then deliver a tailored version
Context 2 → A real job description (copy from LinkedIn, company website, etc.)
Role → A senior recruiter from that exact company type
Task → Tell me what to emphasise, what to add, what to change — then deliver a tailored version
The Job Matcher Prompt
[ROLE]
You are a senior talent acquisition specialist who has hired for [target company type —
e.g., "fast-growing fintech startups"] for 8 years. You know exactly what these
companies look for and how their hiring managers think. You are also an expert at
tailoring resumes to specific job descriptions without fabricating experience.
[TASK]
Compare my resume against the job description below. Then:
1. Identify the match score between my background and this role
2. List what I have that they need (strengths)
3. List what they need that I am missing (gaps)
4. Tell me exactly which 3 bullets to lead with for this role
5. Rewrite my summary section specifically for this job
6. Deliver a fully tailored version of my resume
[MY RESUME — UPDATED VERSION]
[PASTE YOUR REWRITTEN RESUME FROM STEP 3 HERE]
[JOB DESCRIPTION]
[PASTE THE FULL JOB DESCRIPTION HERE — copy from LinkedIn or company site]
[COMPANY/ROLE INFO]
Company: [Company name if known]
Role: [Exact job title]
Seniority: [Junior / Mid / Senior / Lead]
[OUTPUT FORMAT]
## MATCH ANALYSIS
Match Score: X/100
What I have that they need:
✓ [skill/experience] — "they said [quote from JD], I have [quote from resume]"
✓ [skill/experience] — ...
[list all matches]
What they need that I am missing:
✗ [requirement] — how serious is this gap? [Low/Medium/High]
✗ [requirement] — ...
## TOP 3 BULLETS TO LEAD WITH FOR THIS ROLE
1. "[bullet]" — why: [why this one resonates for this specific role]
2. "[bullet]"
3. "[bullet]"
## REWRITTEN SUMMARY FOR THIS ROLE
[New 2-sentence summary, written to match the JD language and priorities]
## FULLY TAILORED RESUME
[Deliver my complete resume reordered and adjusted for this specific role — same
experience, different emphasis, JD keywords integrated naturally]
Pro tip: Run this Job
Matcher prompt every time you apply for a new role. Paste in a different job description
each time. 20 minutes per application — but now you send a genuinely tailored resume,
not a generic one. This single prompt can meaningfully increase your interview rate.
I found a real job description to test this with
I ran the Job Matcher with both my resume and the job
description pasted in
I received a match score, gap analysis, and tailored resume
5
Build Your HTML Dashboard
Prompt Claude to build a visual interface for your whole resume
system
Locked
This is where it gets insane: You are about to build a fully interactive resume optimizer dashboard that runs live inside Claude — as a Claude Artifact. No files to save. No browser to open. Your dashboard appears right inside the conversation, and you can use it immediately. This is how real AI builders work.
⚡ What is a Claude Artifact? When you ask Claude to build an interactive tool, it can render HTML/CSS/JavaScript directly inside the conversation as a live, clickable application. You will see a panel appear next to the chat — that is your dashboard. You can interact with it, iterate on it, and keep improving it conversationally.
First: Set Up a Claude Project (5 minutes)
A Claude Project keeps your resume permanently loaded — so every prompt in this project automatically has your resume as context. No more copy-pasting.
1.Go to claude.ai → click "Projects" in the left sidebar
2.Click "Create Project" → name it
AI Resume Optimizer3.In the project, click "Add content" → paste your Resume Content Block from Step 1
4.In "Project instructions", paste the system prompt below ↓
You are my personal AI Resume Optimization System. You have my full resume loaded in this project.
When I ask you to run a tool, use the resume data from the project context — I should never need to paste it again.
You have 3 tools available:
1. ATS SCANNER — Analyse my resume against a target role and score it 0-100
2. BULLET REWRITER — Critique, rewrite, and score my bullet points (3-step chain)
3. JOB MATCHER — Compare my resume against a specific job description
Always be direct, specific, and honest. Never inflate scores. Use the exact output formats I specify.
When I ask you to "build the dashboard", create it as an interactive Artifact with all 3 tools accessible from a tabbed interface.
The Dashboard Artifact Prompt
Send this prompt inside your Claude Project. Claude will generate a live interactive dashboard as an Artifact — it appears in a panel next to the chat. You can click tabs, paste text, and use it immediately. Notice how this prompt combines all 4 techniques you learned.
[ROLE]
You are a senior frontend developer who specialises in building clean, single-file
developer tools. You write elegant HTML/CSS/JavaScript without any frameworks or npm.
You prioritise function, clarity, and dark mode aesthetics.
[TASK]
Build me an interactive resume optimizer dashboard as a Claude Artifact. This should render as a live, usable tool right here in our conversation.
[CONTEXT — THE 3 TOOLS IT MUST CONTAIN]
The dashboard has 3 tabs:
TAB 1: ATS SCANNER
- A large textarea where I paste my resume text
- A dropdown to select target role category (Software Engineering, Data / Analytics,
Product Management, Marketing / Growth, Finance / Consulting, Pharma / Life Sciences,
Other)
- A "Run ATS Scan" button
- Below the button: a results panel that shows:
• ATS Score (0-100) as a large number with a colour-coded bar (red <50, amber 50-75,
green>75)
• Keywords found (green chips)
• Missing keywords (red chips)
• Format issues (warning list)
• Top 3 improvements (numbered list)
- Important: The tool does NOT call any API. The "results" area shows pre-written
guidance based on common ATS patterns, with a note to paste results from Claude.
TAB 2: BULLET REWRITER
- Input: paste original bullet points (one per line)
- Target role input (text field)
- Three sub-tabs: Prompt A (Critic) / Prompt B (Rewriter) / Prompt C (Scorer)
- Each sub-tab shows: the prompt text pre-filled (copyable), a "Copy Prompt" button,
and a notes area for pasting Claude's output
- A side-by-side comparison view: original bullets on left, rewritten on right (user
fills in right side manually)
TAB 3: JOB MATCHER
- Input area 1: "Paste your resume"
- Input area 2: "Paste job description"
- Job title + company fields
- "Generate Match Prompt" button → displays the complete formatted Job Matcher
prompt (pre-filled with whatever was pasted) → big "Copy" button
- Results area: user pastes Claude's match analysis output here to save it
[DESIGN REQUIREMENTS]
- Dark mode only: background #06080b, cards #0c1015, accent colour #10d9a0
- Font: use Google Fonts — DM Sans for body, Syne for headings (import from CDN)
- Tab navigation: clean minimal tabs at the top
- Mobile responsive: works on tablet and desktop (min 768px)
- Status indicators: show which tabs are "complete" (green dot) once user marks them
done
- A header with: "AI Resume Optimizer" title, a small "Project 2 — ASJ AI Builder"
badge
[TECHNICAL RULES]
- Render as a Claude Artifact (interactive HTML)
- No external API calls — this is a prompt-management tool
- All state in memory — clean, self-contained
- Must be fully interactive: tabs clickable, textareas editable, buttons functional
Deliver as a single Artifact. Make it immediately usable. No explanations outside the artifact.
Iteration prompts — improve your Artifact
After the Artifact appears, send these follow-up prompts in the same conversation to upgrade specific parts:
Improve the ATS tab scoring visual
Fix ONLY
the ATS Score display in Tab 1.
The score number should be a large circular progress ring (SVG-based), not a
bar. Green if >75, amber 50-75, red below 50. Keep all other tabs identical.
Add a download-as-PDF button
Add a
"Download Results as PDF" button to each tab. Use window.print() with a @media
print CSS block that makes the active tab content print cleanly. No libraries
needed.
I created a Claude Project and added my resume as project content
I sent the Dashboard Artifact prompt and Claude rendered a live interactive dashboard
All 3 tabs work inside the Artifact — I can click and interact with them
Prompt Playground — Build Live
Use this interactive builder to construct and copy your prompts in real time. Select a
tool, fill in your details, and get a complete ready-to-run prompt. No need to edit a template
manually.
Resume Prompt Builder
Interactive
Which tool are you building?
Your target role
Your resume (paste key bullets — just a few lines to test)
Generated prompt — copy and run in Claude
Select a tool above to generate your prompt...
→ Then paste in claude.ai
Practice mode: Use this playground to
experiment with different role targets. Try the ATS Scanner with a pharma role target vs a data
analyst role — see how the missing keyword list changes. Each role type has different ATS criteria.
Resume Quick Check FORMAT VALIDATOR
Paste your resume text below for an instant formatting pre-flight check. This catches structural issues before you send it to Claude for deep analysis. JavaScript-powered — no AI needed for these checks.
ATS Score Simulator
Based on common ATS patterns,
here is how different resume issues affect your score. Understand the scoring logic before you
run your real scan.
+25
Keyword density
+20
Format compatibility
+20
Section structure
+15
Quantified results
+10
Job title match
+10
Education match
Takeaway: Keyword density alone accounts
for 25 points. That is why the ATS Scanner's missing keyword list is the single most important
output of your analysis — adding 5-8 missing keywords to your resume can jump your score by
10-15 points immediately.
Debug & Fix Common Problems
When Claude gives you a disappointing output from your resume prompts, it's almost
always fixable with a precise follow-up. Here are the 6 most common failures and the exact prompts
to resolve each one.
Problem 01
ATS score is too high or too low — feels wrong
Claude gave you a 90/100
for a clearly weak resume, or 30/100 for a genuinely good one.
Your ATS score seems [too high/too low] given my resume. Re-evaluate
using stricter criteria.
For the re-evaluation:
- Compare against an entry-level [target role] at a competitive company (e.g.,
Goldman Sachs, McKinsey, Google, Pfizer — depending on my field)
- Score against the following standard: a score of 70 means "this would pass initial
ATS filtering at 60% of companies"
- Be more specific about which exact keywords are missing and why each one matters
for this role
Provide a revised score with reasoning. Show me the 3 keywords that, if added, would
have the biggest single impact on my score.
Problem 02
Rewritten bullets sound robotic or AI-generated
The new bullets are
technically correct but don't sound like a real person wrote them.
The rewritten bullets sound AI-generated and overly formal. Rewrite
them again with these constraints:
1. Sound like a real person who is proud of their work — not a recruiter's form
letter
2. Vary sentence rhythm — not every bullet follows the exact same [Verb + metric]
structure
3. One bullet can be slightly longer (up to 25 words) if the complexity of the work
deserves it
4. Avoid these overused phrases: "leveraged", "spearheaded", "synergized",
"utilized", "facilitated"
5. At least 2 bullets should include a specific tool, system, or method by name
Keep the quantified results. Keep the strong verbs. Just make it sound human.
Problem 03
Job Matcher match score seems generic
The match analysis doesn't
reference specific lines from the job description.
Your match analysis was too generic. Redo the matching with a
different approach:
For every match and gap you identify, you MUST quote:
1. The exact phrase from the job description that you are matching
2. The exact bullet or line from my resume that matches it (or doesn't)
Format:
JD says: "[exact quote from job description]"
My resume has: "[exact quote from resume, or MISSING if not present]"
Assessment: [Strong match / Weak match / Gap]
This level of specificity is what makes the tailoring actually useful. Generic
observations like "you have Python experience" are not enough — quote the source
every time.
Problem 04
Tailored resume changes my experience —
fabrication
Claude rewrote bullets that
claim experience or tools you don't have.
Some of the tailored bullets you wrote contain claims I cannot stand
behind — you invented [describe what was fabricated, e.g., "AWS experience I don't
have", "a specific metric I never measured"].
Redo the tailoring with this strict rule:
NEVER add a skill, tool, technology, metric, or responsibility that did not exist in
my original resume.
The only changes allowed:
1. Reorder bullets to lead with the most relevant ones for this JD
2. Add keywords from the JD naturally into existing bullet context, where they
genuinely fit
3. Adjust the summary section to use language from the JD
4. Highlight transferable skills that I already listed
If there is a hard gap (I genuinely don't have something they need), flag it clearly
but do not fabricate experience to cover it.
Problem 05
Chain prompt B forgot the context from Prompt A
Claude's Prompt B rewrite
didn't use the critique from Prompt A — it just rewrote generically.
Your rewrite in the previous message did not use the specific
critiques you identified. Go back and redo it.
For each bullet, I want you to:
1. Quote the specific weakness you identified in your critique (e.g., "no
quantification", "weak verb")
2. Show how your rewrite directly addresses that specific weakness
This makes the chain explicit — each rewrite must be a direct response to a specific
criticism, not a generic improvement.
Start with the bullet you scored lowest in Prompt A and work upward.
Problem 06
Dashboard HTML is missing a tab or broken on
mobile
The dashboard Claude built
is missing Tab 2, or the layout breaks on smaller screens.
The dashboard has a problem: [describe exactly — e.g., "Tab 2 (Bullet
Rewriter) is missing entirely" / "the layout breaks below 768px width — the textarea
overflows the container"]
Fix ONLY this issue. Deliver the complete corrected .html file from <!DOCTYPE
html> to </html>.
Do not truncate or summarise any section. Include every line of the file.
After fixing, check: does every tab render correctly? Does the layout hold at 768px
and 1280px? Are all 3 "Copy" buttons functional?
The debugging mindset: When a prompt
gives a bad output, do not start a new chat and hope for a different result. That rarely works.
Instead, stay in the same conversation and send a precise correction — tell Claude exactly what was
wrong, what rule it violated, and what you want it to do differently. Specificity is always the fix.
Deploy & Document Your System
Your resume optimizer system is 5 prompt files + 1 HTML dashboard. Time to put it on
GitHub, document it properly, and write your LinkedIn post. This is how it becomes a CV project.
What to upload to GitHub
Create a GitHub repository
called
ai-resume-optimizer.
Upload these files:
ats-scanner.md
Your ATS Scanner prompt file (from Step 2)
rewriter-chain.md
All 3 chain prompts (A, B, C) in one file
with section headers
job-matcher.md
Your Job Matcher prompt (from Step 4)
claude-project-setup.md
Claude Project system prompt + Artifact dashboard instructions
README.md
Project documentation — prompt Claude to
write this (below)
Generate Your README.md
Run this prompt in a fresh
Claude conversation to generate your professional project README:
Write a professional README.md for a GitHub repository called "AI Resume
Optimizer".
Context:
- This is a project from my AI builder portfolio (ASJ AI Systems Builder program)
- I built 3 AI-powered resume tools using Claude and prompt engineering
- Tools: ATS Scanner, Bullet Point Rewriter (3-step chain prompt), Job Description Matcher
- All tools use prompt files — no code, no API keys, no server needed
- Also includes an HTML dashboard that manages all 3 tools from a clean UI
- Built using: Structured prompting, Chain prompting, Context injection, Role prompting
README must include:
1. A clear title and 1-paragraph project description
2. "What This Does" — 3 bullet points, one per tool, with a clear outcome for each
3. "How It Works" — brief explanation of the 4 prompting techniques used (technical readers will
appreciate this)
4. "How to Use" — step-by-step: where to find the prompt files, how to run them in Claude, how
to open the HTML dashboard
5. "Prompting Techniques Demonstrated" — a clean table showing: Technique | Where Used | Why It
Was Chosen
6. "Built With" section: Claude, HTML/CSS/JavaScript, Prompt Engineering
7. A short closing line that invites people to use or fork the repository
Tone: professional but accessible. This README should make a recruiter or fellow developer
immediately understand what was built and respect the technical thinking behind it.
Format: valid Markdown. No decorative headers — just clean structure.
Your LinkedIn Post (AI-Written)
This prompt generates a
LinkedIn post announcing your project. Run it, customise the output, and post it today:
Write me a LinkedIn post announcing that I just built an AI-powered Resume
Optimizer using Claude and prompt engineering.
Context:
- I am a [your role/student status] in [your field]
- I built 3 tools: an ATS Scanner, a Bullet Point Rewriter using chain prompting, and a Job
Description Matcher
- All tools are pure prompt files — no code, no API — just carefully engineered prompts
- I also built an HTML dashboard to run all 3 tools from a clean UI
- GitHub: [YOUR GITHUB LINK]
- I built this as part of the ASJ AI Systems Builder program
Post requirements:
- Open with a hook about the job hunt problem this solves — not "I'm excited to share"
- Mention chain prompting specifically — most people have never heard of it, and it's genuinely
impressive
- Include a brief "what it does" that makes people want to try it
- Invite people to take the GitHub link and use it themselves (drives engagement + reach)
- 3 relevant hashtags
- Tone: confident, generous, not corporate
- Length: 180-220 words
Final Quiz — 8 Questions
Test your understanding of structured prompting, chain prompting, context injection,
and role prompting. Score 5/8 or above to unlock your certificate.
Score:
0
/ 8
Your Certificate
Complete the requirements below to unlock your Module 3 certificate from ASJPrompts & Studio · Claude
Skill Engineering Bootcamp.
Steps Completed
0/5
Quiz Score
0/8
XP Earned
0
Certificate
Locked
Submit Your GitHub Repository
Your certificate is tied to a
real project you actually shipped. Paste your GitHub repo URL below to verify and unlock.