← ClaudeAtlas

resume-screen-doctorlisted

Simulate how an ATS/AI screener reads a resume against a specific job description, pinpoint the exact lines that get eliminated, and produce concrete rewrites — a screening diagnosis, not a resume builder. Use this skill whenever the user wants to know why their resume fails screening, or says any of these — "简历", "求职", "投递没回音", "投了没人理", "过不了初筛", "简历过不了机器", "简历诊断", "帮我看看简历", "这个岗位我能投吗", "JD匹配", "简历被刷", "resume screening", "ATS check", "why my resume gets rejected". Also trigger when the user pastes a resume against a JD — even without explicitly asking for a diagnosis.
f-tiger/zaoqi-blueprints · ★ 3 · AI & Automation · score 69
Install: claude install-skill f-tiger/zaoqi-blueprints
# 简历过筛诊断器 · Resume Screen Doctor > 造器图纸 #12 · 诊断类工具 · 难度 ★★☆ > HR 的 AI 在筛你,你的 AI 应该先筛它。 This skill turns the agent into a resume screening diagnostician: it simulates how an ATS / AI screener would read a resume against a specific job description, pinpoints the exact lines that get the candidate eliminated, and produces concrete rewrites. As a final step it can **forge the diagnosis into a standalone shareable HTML tool** (造器模式), so the user walks away with their own tool, not just an answer. ## Workflow Follow these four stages in order. Do not skip Stage 1. ### Stage 1 — Collect the two inputs You need BOTH of these before diagnosing: 1. **JD (职位描述)** — the actual posting text. If missing, ask for it. If the user has no specific JD, ask for the target role + seniority + industry and state that the diagnosis will use a typical JD for that role. 2. **Resume (简历全文)** — pasted text or an uploaded file. If a file is uploaded, read it first. If the user pasted only a resume with no request attached, assume they want a screening diagnosis and confirm the target role in one short question. ### Stage 2 — Simulate the screening Read `references/screening-rubric.md` for the full scoring rubric, then evaluate the resume the way a screening pipeline actually does, in this order: 1. **硬性过滤 (Hard filters)** — degree, years of experience, location, certifications, visa/work authorization if mentioned in the JD. Any hard-filter miss = instant elimination, flag it f