Last Verified: October 28, 2026 — read the skill’s source (v1.0.0) and ran it ourselves on a sample resume
You apply, you never hear back, and nobody tells you why. hiring-panel is a free Claude Code skill that tells you first: an automated screen and a hiring manager try to reject your resume, five more readers judge it alone as five different people, and then it rebuilds your resume with every mistake fixed — same facts, nothing invented.
Below: install it, check it’s in place, run it on a real resume, read the verdict, and the exact before/after from a run we did ourselves.
In this guide: Quick Start · Step 1 check the install · Step 2 run your first panel · Step 3 read the verdict · Step 4 read the rebuilt resume · how the panel runs · fixes · FAQ
Quick Start
- What it is: a skill folder that adds a
/hiring-panelcommand to Claude Code. No app, no website. - Who made it: Alex Chen (@nocodealex), an independent developer. Not made by Anthropic.
- Cost: free and MIT-licensed. No signup, no API key. One run is 8 sub-agent calls on your normal Claude plan.
- You need: Claude Code, Git, and a resume (file path or pasted text).
Paste into your terminal:
git clone https://github.com/alexyc9381/hiring-panel-skill
cp -r hiring-panel-skill/skills/hiring-panel ~/.claude/skills/Rather let Claude do it?
Paste this into Claude Code instead and it installs and checks the skill for you:
https://github.com/alexyc9381/hiring-panel-skill
Install this skill, then confirm /hiring-panel works.Other ways to install
As a plugin: /plugin marketplace add alexyc9381/hiring-panel-skill then /plugin install hiring-panel-skill@hiring-panel-skill. Or copy skills/hiring-panel into ~/.claude/skills/ yourself. Open a new Claude Code session after any of these before the command shows up.
On this page (8 sections)
Step 1. Check the Install
1a. The skill files are in place
~1 min
ls ~/.claude/skills/hiring-panelYou should see one file: SKILL.md.
Then type / in Claude Code: /hiring-panel should be in the list. If it isn’t, exit and start a new session.
1b. Have a resume ready
~1 min
A PDF, a .docx, or just pasted text all work — the skill reads whatever you give it. Having the job posting too (paste it after the resume) makes every reviewer sharper, since the automated screen checks your resume against that posting’s actual keywords.
Step 2. Run Your First Panel
2a. Turn on accept-edits mode
~5 sec
Press Shift+Tab until the status line says accept edits on. The skill writes two files for you at the end; without this it asks you to approve each one.
2b. Give it the resume and the role
~1 min
Everything after /hiring-panel is read as the resume, so put the role or job posting on a new line or right after the file path:
/hiring-panel ~/Documents/resume.pdf
Senior product designer at a fintech startupNo job posting at all? The skill asks you one question first (what job are you applying for), then continues. It worked if Claude says something like running the automated screen, then the hiring manager, then five readers before the checks start.
8 sub-agent calls on your plan
The automated screen, the hiring manager, and five readers are each a separate sub-agent call (7 so far), plus a second automated-screen pass on the rebuilt resume at the end (8 total). All of it counts toward your normal Claude usage — there’s no quick/lite mode.
Step 3. Read the Verdict
When it finishes you get two files in .hiring-panel/<date>-<name>/: PANEL.md (all seven verdicts plus the merged mistake list) and RESUME-IMPROVED.md (the rebuilt resume). Here’s the real PANEL.md from a run we did on a synthetic sample resume (fictional, no real person) applying for a Marketing Coordinator role:
Hiring manager (interview decision): NO
Five readers, alone: PASS · PASS · PASS · PASS · PASS
Unanimous: 7 of 7 reviewers rejected.
#1 mistake (flagged by all 7): zero metrics anywhere — every bullet is a duty, not a result.
After rebuild, automated screen: REJECT → PASS
The real top of that merged list, worst first — this is the part worth reading closely, because it tells you exactly what to fix even before you see the rewrite:
| # | Mistake | Flagged by |
|---|---|---|
| 1 | Zero metrics anywhere — every bullet is a duty, not a result | all 7 |
| 2 | Objective line is generic filler, wastes the top slot | 4 of 7 |
| 3 | Passive/hedging verbs undercut ownership (“helping with”, “assisted”) | 4 of 7 |
| 4 | Skills section is soft-skill filler, missing role-specific tools | 2 of 7 |
| 5 | No named social channels/platforms | 2 of 7 |
From .hiring-panel/2026-10-28-jordan-ellis/PANEL.md, a run we did ourselves on a fictional sample resume using the hiring-panel skill by Alex Chen (@nocodealex), MIT.
- Unanimous reject, like this one? The merged list is your to-do, worst first — fix the top 2-3 before anything else.
- A close or mixed vote? Read which reviewer still said HIRE or MAYBE and why — that’s what’s already working, keep it.
A second opinion, not a guarantee
Seven language models reading only your resume (and the posting, if you gave one) is a structured second opinion, not career or legal advice, and it doesn’t guarantee an interview.
Step 4. Read the Rebuilt Resume
The skill never invents employers, titles, dates, or numbers. Where a fix needs a fact it doesn’t have, it leaves [ADD: ...] right in the resume instead of making something up. Here’s the real opening of the rebuilt version from the same run:
Marketing Coordinator candidate with 2+ years across social media content,
email campaigns, and event support at a retail brand…
[ADD: one line naming the channels/tools you’re strongest in]
Marketing Assistant — Northfield Retail Group (2022 – Present)
– Created and scheduled social media posts and marketing materials across
[ADD: which platforms — Instagram/Facebook/TikTok?]
Those [ADD: ...] tags are a feature, not a bug: they’re the exact questions to answer with your own numbers before you send the resume out. Fill them in, and the structure, verbs, and framing are already fixed.
How the Panel Runs
| Stage | Who | Calls |
|---|---|---|
| Automated screen | Checks readability, titles/dates, and keyword match against the job posting | 1 |
| Hiring manager | Decides interview or not, what decided it in the first 10 seconds | 1 |
| Five readers, alone | Each judges blind, from a different seat (see below) | 5 |
| Rebuild + re-screen | Claude merges all seven verdicts, rewrites the resume, re-runs the automated screen on the result | 1 |
The five readers never see each other’s notes, and each answers in under 120 words:
The hiring manager (day-one lens) · Another applicant (rival) · The CEO (impact lens) · The recruiter (30 seconds, top third only) · The teammate (how you work with others)
A merge step (done by the main Claude session, not a sub-agent) combines duplicates across all seven verdicts into one ranked list before the rewrite starts.
If Something Goes Wrong
| You see | Why | Fix |
|---|---|---|
/hiring-panel isn’t in the command list | The session started before the install, or the folder is in the wrong place | Check Step 1a, then exit Claude Code and start a new session |
| An approval prompt for every file | The default permission mode asks before each write | Shift+Tab to accept-edits mode before you start (Step 2a) |
| It asks you what job you’re applying for | No job posting or role was given after the resume | Expected — answer once and it continues |
| Readers argue about facts you never gave | Each reader only sees the resume (and the job, if given), not your chat | Make sure the facts you want judged are actually written on the resume |
Lots of [ADD: ...] tags in the rebuild | The skill won’t invent numbers it doesn’t have | Fill them in with your real metrics, then the resume is done |
| You hit your usage limit mid-run | 8 sub-agent calls in one go, no lite mode | Wait for your limit to reset, or run it on a quieter resume first to confirm the setup works |
To remove it: rm -rf ~/.claude/skills/hiring-panel
Official Sources
- alexyc9381/hiring-panel-skill on GitHub (README, source, license)
- The author’s own step-by-step guide at chen.media
FAQ
Is it really free?
The skill is free and MIT-licensed with no signup or API key. The run itself uses your Claude Code plan: one full panel is 8 sub-agent calls.
Does it touch my other files?
No. It only writes into .hiring-panel/<date>-<name>/ in the folder you run it from.
Can it be wrong?
Yes. The reviewers are language models reading only your resume and the posting you gave them. A thin resume gets a thin verdict — the value is in the specific, quoted mistakes, not the vote count alone.
Why five separate readers instead of one more detailed review?
One Claude giving “detailed feedback” still sees its own earlier framing. Here, five sub-agents each judge blind as a different person, so the same objection showing up from more than one reader means it’s a real pattern, not one model’s opinion.
Related tools
Also free: our /court skill guide (12 Claudes put an idea on trial) and 160+ subagents guide.
/hiring-panel skill by Alex Chen (@nocodealex): github.com/alexyc9381/hiring-panel-skill