Resume

Resume

Purpose

Write a resume that survives a six-second scan and a hiring manager's skepticism. Most resumes fail on the same thing: they list responsibilities rather than demonstrating outcomes.

When to Use

• Writing or rewriting a resume.
• Tailoring a resume to a specific role.
• A resume that generates no responses.
• Reviewing someone else's resume.

Capabilities

• Structure and prioritization.
• Turning responsibilities into evidence.
• Tailoring without dishonesty.
• Handling gaps, career changes, and short tenures.
• Applicant-tracking-system considerations.

Inputs

• The person's actual experience, in detail.
• The specific role, and its posting.
• What is genuinely relevant to it.

Outputs

• A resume where each bullet is evidence.
• A version tailored to the role, drawing on real experience.
• No adjectives doing the work that numbers should do.

Workflow

1. Extract the outcomes — For each role: what changed because you were there? Not what you were responsible for. What was different afterwards.
2. Quantify — Numbers survive a six-second scan; adjectives do not. If a number is not available, use a comparison, a scale, or a before-and-after.
3. Lead with the strongest evidence — The top third of the first page is nearly all that is read on the first pass. Put the best thing there.
4. Tailor to the posting, honestly — Reorder, re-emphasize, and use the posting's vocabulary where it genuinely describes what you did. Never invent.
5. Cut everything irrelevant — A resume is an argument for one role, not a complete autobiography.
6. Check it survives a keyword scan — Most applications pass through automated filtering. If the posting says "Kubernetes" and your resume says "container orchestration", you may be filtered out for a skill you have.

Best Practices

• "Responsible for the deployment pipeline" says nothing. "Cut deploy time from 45 to 6 minutes; deploy frequency tripled" says everything. The first describes a job description; the second describes a person.
• Adjectives are what you write when you do not have evidence. "Highly motivated", "results-driven", "passionate" — cut all of them, always. They are free to write, so they carry no information.
• Numbers do not have to be impressive to be useful. "Reduced the on-call page volume from ~30 to ~8 per week" is a small number and a compelling one.
• Tailoring is reordering and re-emphasizing what is true. It is not inventing. The interview will find the invention.
• Two pages maximum, and one page for under ten years of experience. Length is not evidence of substance.
• If a hiring manager can substitute anyone else's name at the top and the resume still reads correctly, it is not a resume — it is a job description.

Examples

The single most important transformation:

BEFORE — a job description. Anyone in the role could have written this.

  Senior Software Engineer, Acme Corp (2022-2026)
  - Responsible for backend services and API development
  - Worked with cross-functional teams to deliver features
  - Participated in code reviews and mentored junior developers
  - Utilized Python, PostgreSQL, and AWS in a microservices architecture

AFTER — evidence. Only this person could have written this.

  Senior Software Engineer, Acme Corp (2022-2026)
  - Cut checkout p99 latency from 3.2s to 340ms by replacing an N+1 query
    pattern in the pricing service. Checkout abandonment fell 4 points.
  - Led the migration of 12 services off a shared database, eliminating the
    incident class that caused 3 of the year's 6 SEV-1s.
  - Built the deploy gate that made integration tests mandatory. Production
    incidents caused by deploys fell from ~2/month to 0 over 8 months.
  - Mentored 3 engineers; 2 promoted to senior within 18 months.

  Python, PostgreSQL, AWS (ECS, RDS, SQS), Terraform

Same job. The first version could describe ten thousand people. The second describes one, and every line is checkable.

Tailoring honestly:

The posting emphasizes: reliability, incident response, Kubernetes.

The candidate has all three, but their resume leads with feature work, because
that is what they spent the most time on.

Tailor by reordering — not by inventing:
  - Move the incident-reduction bullet to the top of the most recent role.
  - Move the database migration bullet (which eliminated an incident class)
    to second.
  - The feature work moves down. It is still true, still there, and no longer
    the first thing read.
  - The posting says "Kubernetes"; the resume said "container orchestration".
    The candidate used Kubernetes. Say "Kubernetes" — an automated filter does
    not do synonyms.

Nothing was invented. The same experience is presented in the order that
answers the question the reader is actually asking.

Notes

• The six-second scan is real, and it lands on the top third of page one. Whatever is there is your resume; the rest is read only if that part worked.
• Automated keyword filtering is imperfect and literal. Using the posting's exact terminology for skills you genuinely have is not gaming the system; it is not being filtered out for a synonym.
• If a bullet does not contain a number, a scale, or a specific outcome, ask what it is doing there. Usually the answer is that it is padding a role that had one real achievement, and it would be stronger alone.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: nimadorostkar
• Source: nimadorostkar/Claude-Skills-collection
• License: MIT
• Homepage: https://github.com/nimadorostkar/Claude-Skills-collection

Install and usage instructions live in the source repository linked above.