resume-polish-craft
Skill1 starsChanged 36 days agoMITView on GitHub
Polish a resume that already has the right content - compress page count, tighten bullet lead-ins, decide which sentences get bolded, keep the persona consistent with the target function, and diagnose layout before deleting anything. Triggers on "压缩成一页", "这条引导词改四个字", "哪些句子要加粗", "title 写得有点逼格", "怎么缩不进去". Not for matching a resume against a JD (resume-jd-matcher) and not for turning one project into an entry (vibecoding-resume-story).
In this repository30 other files it gives its agents.
- anna-app-publishSkill
- campus-jobs-radarSkill
- career-pivot-researchSkill
- cf-pages-deploySkill
- company-risk-checkSkill
- github-repo-launch-polishSkill
- github-repo-repair-auditSkill
- headless-webapp-screenshotsSkill
- historical-remotion-filmSkill
- interview-debriefSkill
- job-application-screenerSkill
- job-application-trackerSkill
- llm-output-eval-gateSkill
- memory-sync-abcSkill
- offer-decision-frameworkSkill
- online-assessment-solverSkill
- personal-secretary-memorySkill
- personal-tone-writingSkill
- portfolio-github-workflowSkill
- recruiting-source-huntingSkill
- remotion-delivery-qaSkill
- resume-jd-matcherSkill
- resume-jd-tailorSkill
- session-checkpointSkill
- session-log-archaeologySkill
- single-file-html-toolSkill
- static-site-deploy-verifySkill
- vibecoding-resume-storySkill
- video-portfolio-performanceSkill
- workflow-distillerSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
When in doubt, ask one sentence.
- Distinctive: line 34
Test: no rhetoric, spoken aloud it is plain speech, at a glance you know what this line covers.
- Very specific: line 40
Delete the tech-stack line.
- Distinctive: line 43
Keep Agent / tool selection / prompt engineering — that is AI fluency, a plus for operations.
- Distinctive: line 44
Keep AI coding tools in the skills row (uses AI tools = plus).
- Distinctive: line 61
Not worth bolding: setup sentences, busywork, zero-number process lines, noun piles.
- Distinctive: line 62
One bullet gets exactly one bold: All bold = no bold; the eye has nowhere to land.
- Distinctive: line 69
Delete any number he cannot defend ("我完全不知道" ones) — third-party market data, revenue claims.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Resume Polish Craft
The content is already correct. What is left is density, emphasis, persona, and page count. This skill exists because the failure modes here are social, not editorial: every one of the rules below was learned by doing the wrong thing and getting told off.
Rule 0 — Format tasks and content tasks are different tasks
He asks you to mark things. That means bold, underline, layout. It does not mean delete.
Real case: asked "which sentences are worth bolding", I delivered eleven deletions instead. He read that as destroying work he had already approved. Same message, different verb.
- "标 / 加粗 / 划线" → mark only. Do not attach a deletion proposal.
- "删 / 压缩" → then content is on the table.
- When in doubt, ask one sentence. Never guess by escalating the scope.
Rule 1 — Lead-ins are summaries, not slogans
A bullet lead-in is four characters that say what the line is about, in the same register as
技术栈 / 项目背景 / 项目成果.
- ✅ 工具研发 · 流程搭建 · 场景验证 · 提效方案 · 需求拆解 · 合规优化 · 矩阵搭建
- ❌ 从零造轮 / 深耕运营追求闭环 — invented slogans; reader cannot tell what the line says
- ❌ "卡在出稿,自建工具" — four characters means four characters. Do not upgrade it into a tidy parallel couplet because it reads better. It reads better for you, not for him.
Test: no rhetoric, spoken aloud it is plain speech, at a glance you know what this line covers.
Rule 2 — Persona follows the target function
Applying to operations means the resume must not smell like engineering.
- Delete the tech-stack line. Naming
Python (9 modules / 849 lines)makes an HR screener mislabel the candidate in ten seconds. - "基于 Python 独立开发 Agent" → "独立搭建脚本生成 Agent". No line counts, no module counts.
- Keep Agent / tool selection / prompt engineering — that is AI fluency, a plus for operations.
- Keep AI coding tools in the skills row (uses AI tools = plus); the same name in a tech-stack row reads as "writes code" and misfires.
- Intern boundary: may "参与公司级项目 / 主动开发工具 / 被抽调 / 沉淀 SOP". May not "主导项目 / 负责体系 / 规划战略". Downgrade 负责 to 承担.
- Raising a title on a team project: mark the slice you owned
(
AI 提效侧核心贡献人员). On a solo project use independence words instead (独立主理人).
Rule 3 — Bold is sentence-level, one per bullet
"Which sentences are useful" is a question asked inside a bolding pass. It means: pick the sentences worth emphasising. It does not mean: decide what survives.
Worth bolding, pick one:
- has a number (throughput, pass rate, turnaround, count, views, orders)
- has a method (拆解 27 份爆款 / 横向评测 15+ 款 / 六步调研法)
- has a deliverable (rule library, SOP, agent, adaptation matrix)
Not worth bolding: setup sentences, busywork, zero-number process lines, noun piles. One bullet gets exactly one bold. All bold = no bold; the eye has nowhere to land.
Rule 4 — Number discipline
- A conflict between two figures is stated once, then he decides. Never pick for him, never judge which source is more credible.
- Once he gives a number, write it, even if an older file disagrees.
- Delete any number he cannot defend ("我完全不知道" ones) — third-party market data, revenue claims.
- Proposing a new number (say
15+ 款工具) requires shipping the backup list in the same message, so he can count before committing. - Component figures must sum to the headline figure.
Rule 5 — Diagnose layout before deleting content
Page count over budget does not mean "delete a project". Read the parameters first.
页边距:docx 的 w:sectPr/w:pgMar(twips,567 = 1cm)
字号 :段落 rPr 的 w:sz(21 = 10.5pt)
行距 :段落 pPr 的 w:spacing w:line(320 = 16pt)
行数 :版心高 ÷ 每行高度Usual fix: font size 10 + fixed line spacing 13 pt (floor = font size × 1.2, below that glyphs get clipped) + delete empty paragraphs.
Deletion order, lossless to lossy: empty paragraphs → elective courses → skills laid out in columns → duplicated project background → the student-activities block. Project experience is the hard cargo. Touch it last.
Rule 6 — Mass applications: one page beats two
HR screens in 10-30 seconds and the second page is usually never opened; for a new grad one page is the default expectation and two pages reads as padding. Ship a one-page version for mass applications and keep a two-page version on file for referrals.
Rule 7 — Deliver in the chat, not as a file
- Say it in the conversation. A Markdown file is not read.
- A one-line rule is too vague. Thirty scattered words are too granular. The right unit is one sentence per line, grouped by module.
- Being corrected three times in a row means stop and ask. Do not compensate by doing more.
Self-check before sending
图看全否 / 数字核过否 / 引领句像人话否 / 与下文重复否 / 文件放对位置否。
Red lines
- 🔴 Never delete a field name he approved (技术栈 / 项目背景 / 项目成果). Compress the prose under it, not the label.
- 🔴 Never silently rewrite his file. Mark only; editing the document needs a backup and his yes.
- 🔴 Never expand the verb. Mark stays mark.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

