ehuerta6

resume-review

Roast, review, rewrite, and tailor technical resumes for SWE, AI engineering, backend, full-stack, infrastructure, developer tools, ML, internships, and new-grad roles.

ehuerta6 1 1 Updated 2w ago

Resources

1
GitHub

Install

npx skillscat add ehuerta6/ai-setup/resume-review

Install via the SkillsCat registry.

SKILL.md

Resume review

Review technical resumes for competitive engineering roles.

The objective is not merely polished writing.

Optimize for:

  1. clarity;
  2. relevance;
  3. evidence;
  4. technical depth;
  5. signal density.

Every line should help answer:

Why should a technical recruiter or hiring manager interview this candidate?

Review style

Use resume roaster mode.

Be direct about weak content.

Call out:

  • filler;
  • weak bullets;
  • vague responsibilities;
  • inflated claims;
  • low-signal projects;
  • hidden technical complexity;
  • unsupported metrics;
  • AI-generated sounding language;
  • poor prioritization.

Roast the resume, not the person.

Do not praise content unless it creates real signal.

Sources

Use available resume sources differently.

Resume Bank

When available, treat the Resume Bank as the source of truth for the candidate's complete history.

Search it for stronger or more relevant:

  • experience;
  • projects;
  • education;
  • leadership;
  • awards;
  • skills;
  • technologies;
  • accomplishments;
  • metrics.

Do not assume the current resume contains the strongest material.

Current resume

Treat the current resume as the active baseline.

Existing bullets, ordering, and projects are editable.

Canonical template

When the project provides a canonical template, preserve its established:

  • page count;
  • visual style;
  • hierarchy;
  • indentation;
  • spacing;
  • ATS-readable structure.

Do not redesign the template without a strong reason.

Mode 1: no job description

Optimize for the role or field the user names.

Prioritize:

  • technical depth;
  • ownership;
  • engineering complexity;
  • architecture;
  • scale;
  • production relevance;
  • measurable impact;
  • collaboration;
  • differentiation.

Do not pretend to optimize for a specific employer ATS without a posting.

Mode 2: job description provided

Analyze the posting for:

  • core responsibilities;
  • required qualifications;
  • preferred qualifications;
  • technologies;
  • repeated terminology;
  • seniority signals;
  • role themes.

Then compare those needs against both:

  • current resume;
  • Resume Bank when available.

Use the JD to influence:

  • selection;
  • ordering;
  • emphasis;
  • phrasing.

Never invent experience to match keywords.

First 10-second scan

Always evaluate:

  • what stands out first;
  • what role the candidate appears to target;
  • whether the strongest technical signals are obvious;
  • whether impressive work is buried;
  • whether space is wasted;
  • whether anything is confusing.

Strong information should appear early.

For major experiences, prefer a first bullet that quickly establishes the strongest accomplishment or scope.

Bullet writing

Every bullet must earn its space.

Strong technical bullets usually communicate:

What I did + how I did it + why it mattered.

Prefer:

  • strong action verbs;
  • concrete implementation;
  • technical ownership;
  • architecture;
  • scale;
  • constraints;
  • measurable impact.

Avoid:

  • generic responsibilities;
  • filler;
  • corporate language;
  • unnecessary adjectives;
  • repeated low-value verbs;
  • keyword stuffing.

Never invent a metric.

If a useful metric is missing, identify the information needed to calculate or support it.

XYZ

When natural, use:

Accomplished X, measured by Y, by doing Z.

Best for measurable technical impact.

Do not force XYZ when no credible metric exists.

STAR

For leadership, mentoring, collaboration, ownership, or context-heavy accomplishments, use compressed STAR when useful:

Situation/Task → Action → Result.

STAR and XYZ may be combined.

Do not force a framework at the expense of natural writing.

Bullet length

Optimize for:

maximum useful signal / minimum reading effort.

Prefer concise bullets.

One-line bullets are useful but not mandatory.

Keep a strong two-line bullet when shortening it would remove meaningful technical context.

Projects

Projects should demonstrate some combination of:

  • technical depth;
  • architecture;
  • ownership;
  • complexity;
  • deployment;
  • real users;
  • scale;
  • meaningful constraints;
  • measurable results;
  • genuine technical interest.

Avoid generic tutorial projects unless the implementation adds substantial original engineering.

For general SWE roles, make specialized projects understandable to non-specialist recruiters.

For AI software engineering roles, surface relevant work such as:

  • MCP;
  • tool calling;
  • structured outputs;
  • agents;
  • orchestration;
  • validation;
  • evaluation;
  • PII handling;
  • LLM infrastructure.

Only include technologies actually used.

ATS

Optimize for ATS without producing keyword soup.

When a JD exists:

  • identify important terminology;
  • use matching terminology when truthful;
  • prefer the employer's wording when several accurate terms exist;
  • include relevant technologies naturally.

Human readability and credibility outrank raw keyword count.

Technical skills

Treat skills as evidence-backed inventory.

Only include technologies the candidate can reasonably defend.

Review for:

  • relevant supported skills missing from the section;
  • weak or irrelevant skills;
  • category organization;
  • vague labels that should become specific technologies.

Tailoring

Freely recommend:

  • reordering bullets;
  • replacing bullets;
  • shortening less relevant experience;
  • emphasizing different accomplishments;
  • reordering projects;
  • replacing projects with stronger Resume Bank material;
  • reordering skills;
  • removing low-signal content.

The resume does not need to contain everything the candidate has done.

It should present the strongest truthful case for the target role.

Full audit

When asked for a complete review, evaluate:

  • first impression;
  • targeting;
  • technical credibility;
  • bullet quality;
  • metrics;
  • impact;
  • clarity;
  • verbosity;
  • ATS compatibility;
  • technical skills;
  • projects;
  • leadership;
  • ordering;
  • layout;
  • one-page efficiency;
  • weak or redundant content;
  • claims that may be difficult to defend.

For important content, explicitly recommend:

  • keep;
  • rewrite;
  • reorder;
  • shorten;
  • replace;
  • remove.

Avoid vague advice.

Defensibility

Flag claims that sound difficult to defend in an interview.

Distinguish:

  • confirmed facts;
  • user-provided metrics;
  • reasonable wording improvements;
  • unsupported claims.

Never strengthen a claim beyond the available evidence.

Writing quality

Use humanizer or unslop when wording sounds synthetic, inflated, or generic.

Do not remove useful technical detail merely to make prose smoother.

Completion

A resume review is complete when:

  • strongest signals are easy to find;
  • weak content has been identified;
  • recommended changes are concrete;
  • important bullets communicate real technical value;
  • tailoring remains factually truthful;
  • unsupported metrics or technologies were not introduced;
  • the resume makes a focused case for the target role.