Resume review
A stack of skills for your AI agent, made for people starting from zero. 103 skills and 149 project briefs — load one into Codex, Claude Code, Copilot, Gemini, or Cursor and do the thing.
npx -y skills add protosphinx/sphinxstack --skill resume-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 17 days oldThe repository was created 17 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Review an existing resume line by line. What a screener sees in six seconds, what to cut, what needs evidence. Use when they say "review my resume", "is my resume good", or before they submit anywhere.
SKILL.md
2.9 KB, 653 tokens by cl100k_base, as published. Nobody here has run it
resume-review
Review a resume that already exists, the way a screener and then a skeptical interviewer would. The kindness is in the bluntness: they are about to send this to strangers who will not explain why they said no. Done means every problem you can fix together is fixed, and the ones you cannot are named.
Ground rules
- Verdicts with reasons, line by line. "This bullet says nothing a stranger can check" beats "looks pretty good overall". Never soften a real problem into a suggestion.
- Fixes use only material that is true. Rewording weak phrasing is in scope; inventing a metric to strengthen a bullet is not. When a line needs evidence that does not exist, the verdict is "cut or go build the evidence", not "add a number".
- Fix as you go. Each verdict comes with the repair, made now with their yes — this is a working session, not a report.
The six-second pass
First, read it the way a screener does — six seconds, then verdict. Report exactly what registered: the name, the first section header, maybe two bullets, whether it looked crowded. Then the layout facts that decide the six seconds: over one page, dense walls of text, tiny fonts, columns or graphics that break ATS parsing, an unprofessional email address, a photo. Each is a fix made now.
The line pass
Go bullet by bullet. For each line, one verdict:
- Keep — specific, true, and carries a link or checkable fact.
- Rewrite — a real thing said weakly. "Responsible for social media" hides "ran the club account, three posts a week"; ask what actually happened and reword to the fact. Aspiration verbs ("passionate", "motivated") become evidence-backed action verbs or are removed.
- Needs evidence — a claim with nothing behind it. A skills line no other line supports, a project with no link, an achievement with no artifact. They either supply the evidence now or the line moves to Cut.
- Cut — objectives paragraphs, word-cloud skills, "references available", middle-school entries, anything they could not discuss for two minutes in an interview.
Also check the spine: dates consistent and gap-free or gaps ownable, tenses consistent, zero typos (read it aloud), contact links all working.
Done
- Every line carries a verdict, and every Keep/Rewrite/Cut is applied to the file
- One page, ATS-safe, links tested, exported to fresh PDF
- A short honest list of Needs-evidence gaps that survive — with the fix for each being something to build, not something to phrase (https://sphinxstack.com/skills/)
- If the review removed most of the page, hand off to build-resume and rebuild from the interview up
If they are targeting one specific posting, run tailor-resume next on the newly cleaned file.