Best Interview and Resume Skills for Claude Code

We scored the interview and resume skills in the 765-skill directory. The top of this category isn't won by better formatting or fancier templates — it's won by skills that get the order of operations right.

The argument: when language is the product, agents have maximum leverage

Most domains where people apply Claude Code have a reality check built in. Generated code either compiles and passes tests or it doesn't. A generated migration either runs or it doesn't. The agent can be confidently wrong, and the system will catch it.

Job-hunting materials are different. The resume, the LinkedIn post, the interview answer — the deliverable is the language output. There is no compiler. There is only a human reader with ninety seconds and an ATS filter with a keyword list. This cuts both ways: an agent can produce fluent, plausible, completely hollow documents at scale, which is exactly the failure mode recruiters complain about. But it also means that when a skill is well-designed, the leverage is enormous — a well-structured conversation with an agent can excavate material you'd never write down on your own, then shape it into what the reader actually wants.

The scores in this category reflect that split. The top tier is occupied by skills that solve the material problem. The bottom tier is full of formatting wrappers.

The skill that nails the sequence

interview-me (9.4, by addyosmani of the Google Chrome team) sits at the top of the pool, and its edge is conceptual, not cosmetic. Most resume skills start from the document: paste in your old resume, get a polished variant back. interview-me inverts this — it interviews you first, extracting what you actually did and what you actually want, and only then produces the document. The tagline we kept coming back to while scoring it: it extracts what you actually want, not what you say you want.

That inversion matters because the raw material is the bottleneck, not the prose. Almost everyone underwrites their own experience — "did some perf work" hides a quarter-long latency battle with real numbers. A generative pass over a thin document produces polished thinness. An interview pass before generation produces a document with substance in it. The order is the product.

linkedin-interviewer (9.4, by sergebulaev) applies the same sequence to a different output: it runs a source interview with you, then turns the transcript into LinkedIn post material. Two different authors converging on interview-first as the pattern is the strongest signal in the category.

The shaping layer

Once material exists, a second tier of skills shapes it for specific readers.

resume-tailoring (9.1, by varunr89) is the highest scorer of the conventional resume skills because it treats each job posting as a distinct target rather than a filter to game — reweighting your actual experience against the posting's actual demands. At the mechanical end, resume-ats-optimizer (8.0) handles the robot reading your resume before any human does, and review-resume (8.1) checks PM resumes against ten best practices — useful, bounded, and honest about being a lint pass rather than a rewrite.

The most interesting entry in this tier is hiring-panel (8.4): seven Claude instances acting as a hiring committee, each tearing your resume apart from a different perspective — the skeptical engineer, the recruiter, the bar-raiser, and so on. It's adversarial in a way single-agent review can't be, and it's the closest thing in the pool to a rehearsal for the actual audience. The score is a notch below the leaders because the critique, while multi-voiced, still operates on whatever you feed it — garbage in, seven opinions of garbage out. Pair it with an interview-first skill and the combination is stronger than either alone.

One skill worth watching at the margins: muse-commands (8.3), which implements /resume and /save commands for resuming and saving agent personas mid-session. It's plumbing, not content — but persistent sessions are what make a multi-round interview workflow survivable, and it scores well precisely because it does one mechanical job cleanly. Similarly, interview-script (8.1) borrows the JTBD user-interview tradition for research interviews, a reminder that "interview skills" split into two families with very different audiences.

The caveats

The failure mode of this whole category is delegation of judgment. An agent can make any experience sound excellent, and a resume built purely from generated enthusiasm fails exactly where it matters — the interview room, where the claims meet a person asking follow-ups. The top-scoring skills avoid this by anchoring generation in extracted truth: the interview-me pattern works because the material came out of your own mouth. The lower tier — generic "write me a resume" wrappers — skips that step and produces confident fiction. The scoring spread between those two approaches is wide, and it should be.

Where this leaves us

The lesson from this category generalizes well beyond job hunting: when the deliverable is language, the hard part isn't the writing — it's getting the raw material and the intent out of the human first. interview-me and linkedin-interviewer score 9.4 not because they write better than other skills but because they ask first. Any skill whose output is a document should study that sequence.

The full scored set lives in the Skill123 directory.