20:"$Sreact.view_transition" 21:I[9754,["/_next/static/chunks/2t36f5p4pjzlb.js","/_next/static/chunks/1gyqeidjm5eoa.js","/_next/static/chunks/0_fjnxc4tt_l_.js","/_next/static/chunks/1ymt1shyhmu-4.js","/_next/static/chunks/3kv2zf0i7s2wc.js","/_next/static/chunks/1vs_3g_g75zmk.js","/_next/static/chunks/2g4dh802s8oz3.js","/_next/static/chunks/1bc3f0shukf1j.js","/_next/static/chunks/192kgs9v_t14j.js","/_next/static/chunks/0z6pw0v5c_lx5.js","/_next/static/chunks/08r60e6icmtkc.js","/_next/static/chunks/3trx6c0i7_x-3.js","/_next/static/chunks/40sro5bzp9bi_.js","/_next/static/chunks/1wgf3tk38y6cf.js"],"ProjectDetail"] 22:T4b3b, career-ops —— AI 求职指挥中心

01分流轨道 Triage Rail

把每一个职位，放到一条
人工把关的分流轨道上

career-ops 将任何 AI 编码 CLI（Claude Code、Codex、OpenCode、Antigravity 等）变成一座求职指挥中心：扫描职位门户、以 A-F 评分档折算为 1.0-5.0 分数、生成定制简历、跟踪申请流程。它仅在本地运行，它从不替你提交，最后一步闸杆始终握在你手里。

INCOMING扫描端口入站

GRADEA-F 评分档 1.0-5.0

TAILOR生成定制简历

TRACK跟踪申请流程

G · 合法性检查

停 · 丢弃

npx @santifer/career-ops init npm run doctor /career-ops npm run gemini:eval -- "JD text"

开源 · MIT 许可 · 官网 · GitHub 仓库 · 作者 santifer，主要语言 JavaScript。

02端口入站 Portal Inbound

职位从各个门户涌入

把职位链接或招聘描述（JD）文本粘给 CLI，即触发自动流水线：career-ops 扫描职位门户，将每个职位拉到轨道前端，作为一条待检的报文排队。它支持越来越多职位板提供商，让入站源持续汇集，而不是让你逐个翻找。

扫描入站 → 职位板提供商 remotli getro

03评分档位 Grade Notches

A-F 评分块，折算成一个数

每个职位按结构化 A-F 评分块逐项评估，最终折算为 1.0-5.0 的单一分数档。它不是一句模糊评判，而是一条可复核的评分条：你可以顺着每一档回头核对，为什么它会走到这一格。首次评估需要你先给出个人背景，否则评分会明显偏弱。

ABCDEF → 1.0 - 5.0

04人工闸杆 Manual Lever

闸杆在你手里，G 是安全闸

career-ops 从不替你提交申请，最终决定权始终在你。每个轨道末尾是一根人工闸杆：是否继续、是否按下。而在判分之前，Block G 合法性检查会先拉下安全闸，标记可疑或不合规的职位——这一格既决定职位是否继续，也是轨道上唯一由人力兜底的停车器。

G · 合法性检查通过 → 继续停 → 丢弃

05批量并行轨道 Batch Lane

一条轨道不够时，并行铺开

当你有一堆职位要处理时，子代理可在并行轨道中同时评估 10+ 个职位，把排队时间压下来。你仍然是每一条轨道的最终操作者，每一条都保留人工闸杆。

子代理10+ 并行批量评估

06快速开始 Quick Start

从 npx 一行命令起步

在当前目录跑初始命令，随后在任意 AI 编码 CLI 里用一句话切换模式：让系统把 archetypes 改成后端工程角色、翻译为英文、往 portals.yml 加五家公司，或把你粘贴的简历更新进档案。它依赖已配置的 AI 编码 CLI 环境，且首次评估需要先提供个人背景。

npx @santifer/career-ops init

npm run doctor # 校验所有前置条件

/career-ops # 呼出模式（若 CLI 支持斜杠命令）

07结果与局限 Outcomes & Caveats

以作者的盖章为准

作者声称曾用这套系统评估 740+ 个职位、生成 100+ 份定制简历，并凭它拿下 Head of Applied AI 职位。这只是作者自述，未经独立验证——请把它当作一条盖章标注，而非客观指标。同样请留意：评估首次偏弱、需要个人背景；需要先配好 AI 编码 CLI 环境。

作者声称 740+ 职位评估 · 100+ 定制简历 · Head of Applied AI

23:T49a8, career-ops — AI job search command center

01Triage Rail

Every open role runs a triage rail,
human on the final lever

career-ops turns any AI coding CLI (Claude Code, Codex, OpenCode, Antigravity and more) into a job search command center: it scans job portals, evaluates listings through an A-F rubric into a single 1.0-5.0 score, generates tailored ATS CVs, and tracks applications. It runs locally, it never submits on your behalf, and the last lever stays in your hands.

INCOMINGscan inbound

GRADEA-F rubric 1.0-5.0

TAILORtailored ATS CV

TRACKtrack applications

G · legitimacy check stop · drop

npx @santifer/career-ops init npm run doctor /career-ops npm run gemini:eval -- "JD text"

Open source · MIT licensed · Website · GitHub repository · by santifer · primary language JavaScript.

02Portal Inbound

Offers flood in from every portal

Paste a job URL or JD text to the CLI and it fires the auto-pipeline: career-ops scans job portals and pulls each listing to the front of the rail, queued as a ticket waiting to be reviewed. Newer job board providers keep inbound feeding steadily, so you are not hunting through each board one by one.

scan inbound → job board providers remotli getro

03Grade Notches

A-F scored blocks, folded to one number

Each listing is evaluated against structured A-F scored blocks and folded into a single 1.0-5.0 score notch. It is not a vague verdict but a reviewable grading bar: you can walk back through each notch to see exactly why a role landed there. First evaluations run noticeably weak until you give the system your personal context.

ABCDEF → 1.0 - 5.0

04Manual Lever

The lever is yours; G is the safety break

career-ops never submits an application, and the final call is always yours. Each end of the rail carries a manual lever: continue or drop, pressed only by you. Before any grade, the Block G posting-legitimacy check pulls the safety break, flagging suspicious or non-compliant listings — one notch that decides whether a role keeps moving, and the only stop the system will let a human hold.

G · legitimacy check pass → continue stop → drop

05Batch Lane

One lane is not enough — spread them in parallel

When a pile of roles is waiting, sub-agents evaluate 10+ offers in parallel lanes to cut the queue time. You remain the final operator of every lane, and each one still keeps its manual lever.

sub-agents10+ parallelbatch review

06Quick Start

One npx line to get going

Init the project, then switch modes inside any AI coding CLI with a plain sentence: have it change the archetypes to backend engineering roles, translate the modes to English, add five companies to portals.yml, or update your profile with the CV you are pasting. It relies on a configured AI coding CLI environment, and early evaluations want your personal context first.

npx @santifer/career-ops init

npm run doctor # validates all prerequisites

/career-ops # open the mode (if your CLI supports slash commands)

07Outcomes & Caveats

Take the author's claim as stamped

The author claims to have used the system to evaluate 740+ job offers, generate 100+ tailored CVs, and land a Head of Applied AI role. This is self-reported and not independently verified — treat it as a stamped annotation, not an objective metric. Also note: first evaluations are weak and need your personal context, and you must have an AI coding CLI environment configured.

Author claim 740+ offers evaluated · 100+ tailored CVs · Head of Applied AI

a:[["$","$20",null,{"enter":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-enter","default":"none"},"exit":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-exit","default":"none"},"update":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-update","default":"none"},"default":"none","children":["$","$L21","cmt2zh2qn01hn6271mfg1rbb9:zh-CN:beautified",{"project":{"id":"cmt2zh2qn01hn6271mfg1rbb9","slug":"santifer-career-ops","name":"career-ops","description":"将 AI 编码 CLI 变成求职指挥中心","authorName":"santifer","authorVerification":{"verified":false,"verifiedAt":null},"category":"development_tools","subcategory":"data_ai_tools","experienceUrl":"https://career-ops.org/","qualityScore":100,"featured":false,"officialSite":null,"views":{"original":{"available":true},"official":{"available":false,"reason":"disabled_by_policy"},"beautified":{"available":true,"reason":null},"defaultView":"beautified"},"beautifiedView":{"revisionId":"cmt2zh2r401hz6271hh2zmkak","kind":"artifact","artifact":{"schemaVersion":"localized-static-html-v4","kind":"static_html","locales":["zh-CN","en"],"defaultLocale":"zh-CN","documents":{"zh-CN":"$22","en":"$23"},"designDirections":{"schemaVersion":"dodot-design-directions-v1","directions":[{"id":"direction-a-scorecard","name":"The Desk Scorecard","projectMetaphor":"A candidate's job search running as a single desk scorecard that a human reviewer fills by hand. Every listing is graded on the A-F rubric and collapses into one 1.0-5.0 number; Block G is the extra gate the reviewer must tick before anything proceeds.","signatureVisual":"One oversized \"X.X\" score value computed from six grade blocks (A-F) stacked vertically, each block rendered as a light ruled ledger row; a seventh short block labeled G sits beneath them lifted toward a final pass/fail line.","heroGeometry":"A tall central rule of six horizontal ledger rows (A-F) that narrow as they converge into a single large numeric value, with the legitimacy gate G drawn as a detached tick above the number.","sectionSequence":["lead-scorecard","rubric-blocks","legitimacy-gate","human-decision","pipeline-quickstart","outcomes-caveat","author-lineage"],"surfaceModel":"Paper desk ledger: ruled horizontal hairlines, faint column guides, stamped ticks, and task-paper panels that sit on a warm neutral desk. Panels use hard edges and a light fold/drop rather than glow.","palette":"Ink on paper: near-black #16120e text, cream-paper #f5f1e8 background, ledger accent red #c0392b for grades and gates, with muted green #2f7d4f and steep amber #b7791f for pass/warn states.","typography":"A monospaced slab for the headline scores and grade labels (table/ledger feel) paired with a humanist serif for body, echoing a typed review card.","desktopComposition":"Asymmetric two-stack: the scorecard ledger occupies the left ~58% with a repeated ruled x-ray of the rail; the right column holds the one-line value prop and a real CTA, with the seven-line scan field under it.","mobileTransformation":"The six grade rows reflow into a single stacked tick-sheet, the X.X number moves to the very top, and the scorecard becomes one continuous hand-scroll of ruled lines instead of two columns.","mediaStrategy":"No media region. The scorecard ledger itself is the derived visual; it expresses the verified A-F-to-number relationship, not a product screenshot.","forbiddenPatterns":["three metric cards under a slogan","terminal or CLI shell chrome","decorative glow or particle grid","fake product screenshot or dashboard UI"],"compositionFingerprint":"scorecard-ledger-hairlines-blocks-AF-G-gate-paper-grades"},{"id":"direction-b-triage-rail","name":"The Triage Rail","projectMetaphor":"A job search as a physical triage and sort rail: every offer arrives from some portal, is run through the reviewer's steps (scan, grade, tailor, track), and at each flow the human holds the final lever — the system never pulls it. The Block G legitimacy check is the safety break on the rail.","signatureVisual":"A folded timeline rail running left to right with four stations (INCOMING → GRADE → TAILOR → TRACK) and a fifth break-lever labeled G that, when engaged by a hand, decides whether a job continues or is dropped.","heroGeometry":"A horizontal station rail that reads as a conveyor: small offer tokens entering on the left, passing a central grade dial, then a single over-engineered lever labeled G that pivots toward a decision notch on the right.","sectionSequence":["triage-rail","portal-inbound","grade-notches","manual-lever","batch-lane","quickstart","outcomes-caveat"],"surfaceModel":"Mechanical control panel: a single neutral operational surface with raised station segments, recessed tracks, torque rings, and stamped labels; panels look like machined plate, not glass cards.","palette":"Industrial steel-and-signal: slate graphite #171c22 background, machine gray #f2f4f6 foreground text, signal red #e5484d for the G stop lever, signal green #46a45c and signal amber #e0a02f for pass and warn.","typography":"A condensed grotesque for rail labels mixed with a monospaced typewriter for station readouts, echoing a machine plaque and its printed token.","desktopComposition":"The rail dominates the upper half full-width as a real left-to-right flow; beneath it a narrow strip of four station annotations acts as the value narrative, with the quickstart commands laid out as tags along the rail bottom.","mobileTransformation":"The horizontal rail rotates 90 degrees into a stacked vertical flow where each station is a full row and the lever becomes a big centered toggle the finger would obviously operate; token annotations collapse to per-row captions.","mediaStrategy":"No media region. The rail and its lever are the derived visual, expressing the scan→grade→CV→track pipeline and the human-override gate from evidence.","forbiddenPatterns":["browser or editor shell","generic feature-card grid","particle or glow backdrop used as a visual crutch","fake terminal output as decoration"],"compositionFingerprint":"triage-rail-horizontal-stations-lever-gate-conveyor-control-panel"},{"id":"direction-c-parallel-desk","name":"The Parallel Desk","projectMetaphor":"A job search handled like a batch desk where many candidate offers arrive at once and are reviewed in parallel — the fan of offer cards across a spread desk, with the human at the decision point. The user receives a wide fan of graded options and keeps the final call on every one.","signatureVisual":"A radar-like fan of offer tokens radiating from a single decision hub in the center: each token carries an A-F grade mark, and the hub holds one small 1.0-5.0 score. Every token either stays in the fan or is pulled out by the human.","heroGeometry":"A central decision hub surrounded by a radial constellation of offer makers where about 10 sit in an arc, converging on the hub; the Block G legitimacy check is the orbit ring separating near tokens from dropped ones.","sectionSequence":["decision-hub","batch-fan","grade-marks","orbit-gate","tailor-print","quickstart","outcomes-caveat"],"surfaceModel":"A single spread desk plane with depth layering: tokens sit at staggered depth like cards on a real desk, the hub is inset rather than elevated, using soft drop shadow and overlap to read as depth, not glow cards.","palette":"Daylight workstation: light warm-gray #eef0ee canvas, deep charcoal #1d2024 ink, token amber #bf7116, hub teal #17807a, legitimacy red #cf3f3f used sparingly on the orbit ring.","typography":"A readable humanist sans for the narrative with a small tabular numeral face for the 1.0-5.0 scores and grades on every token, reinforcing the scored-decision spread.","desktopComposition":"Center-right holds the hub and its arc of tokens; left holds the headline, a short line of value and a real CTA; a bottom strip of quickstart commands reads as a real next step rather than a feature grid.","mobileTransformation":"The radial fan flattens into a simple stacked list of tokens where each row shows one score and its grade mark, and the hub becomes a small sticky header; depth becomes vertical scroll order.","mediaStrategy":"No media region. The fan of scored tokens is the derived visual expressing the parallel batch sub-agents and the human final call, not a product screenshot.","forbiddenPatterns":["symmetrical three-card metrics","hero wall of browser/terminal chrome","background grid or particle starspace","fake chart dashboard"],"compositionFingerprint":"parallel-desk-radial-fan-hub-orbit-gate-score-tokens-decision"}],"selectedDirectionId":"direction-b-triage-rail","selectionRationale":"Direction B (triage-rail) best expresses the project's actual mechanism: a verified pipeline from inbound job listings through A-F grading to tailored CV and tracking, with the Block G legitimacy check and the never-auto-submit human-override as the defining, distinctive facts. Its section sequence is driven by evidence (the flow, the gate, the batch lane, quickstart, caveats) rather than a marketing order; its station rail and lever are a project-specific visual idea without fake product or terminal chrome. Direction A is strong but risks reading as a generic report card; Direction C expresses the parallel-batch strength but centers depth layering that is harder to keep evidence-true and can drift toward a dashboard feel. Direction B also lets the author's claimed 740+/100+ numbers and the caveat section appear as stamped annotations on the rail, keeping unsupported claims clearly framed as the author's claim."},"designBrief":{"schemaVersion":"dodot-design-brief-v1","selectedDirectionId":"direction-b-triage-rail","projectMetaphor":"A job search as a physical triage and sort rail: every offer arrives from some portal, is run through the reviewer's steps (scan, grade, tailor, track), and at each flow the human holds the final lever — the system never pulls it. The Block G legitimacy check is the safety break on the rail.","signatureVisual":"A folded timeline rail running left to right with four stations (INCOMING → GRADE → TAILOR → TRACK) and a fifth break-lever labeled G that, when engaged by a hand, decides whether a job continues or is dropped.","compositionFingerprint":"triage-rail-horizontal-stations-lever-gate-conveyor-control-panel","heroGeometry":"A horizontal station rail that reads as a conveyor: small offer tokens entering on the left, passing a central grade dial, then a single over-engineered lever labeled G that pivots toward a decision notch on the right.","sectionSequence":["triage-rail","portal-inbound","grade-notches","manual-lever","batch-lane","quickstart","outcomes-caveat"],"surfaceModel":"Mechanical control panel: a single neutral operational surface with raised station segments, recessed tracks, torque rings, and stamped labels; panels look like machined plate, not glass cards.","palette":"Industrial steel-and-signal: slate graphite #171c22 background, machine gray #f2f4f6 foreground text, signal red #e5484d for the G stop lever, signal green #46a45c and signal amber #e0a02f for pass and warn.","typography":"A condensed grotesque for rail labels mixed with a monospaced typewriter for station readouts, echoing a machine plaque and its printed token.","desktopComposition":"The rail dominates the upper half full-width as a real left-to-right flow; beneath it a narrow strip of four station annotations acts as the value narrative, with the quickstart commands laid out as tags along the rail bottom.","mobileTransformation":"The horizontal rail rotates 90 degrees into a stacked vertical flow where each station is a full row and the lever becomes a big centered toggle the finger would obviously operate; token annotations collapse to per-row captions.","mediaStrategy":"No media region. The rail and its lever are the derived visual, expressing the scan→grade→CV→track pipeline and the human-override gate from evidence.","evidencePlan":["Render the four station names (inbound, grade, tailor, track) directly from the verified pipeline language in the projection description and readme headings.","Mark the Block G posting-legitimacy check as the separate break-lever/stop, sourced from summary evidence claim about Block G and the never-auto-submit fact.","State the parallel batch nature (10+ offers) only as a labeled annotation along the batch-lane section, sourced from the parallel sub-agents evidence claim.","Frame the author's claimed 740+ offers and 100+ tailored CVs as a clearly labeled author-claim stamp, not an independent metric, per the caveat evidence.","Show the quickstart commands (init, doctor, /career-ops modes) as authentic next-step tags only from readme commands, not invented controls."],"localeContract":{"zh-CN":"使用实际管道术语：扫描职位门户、以 A-F 评分块评估并折算为 1.0-5.0 分数、坚持 Block G 合法性检查、生成定制简历、跟踪申请；明确系统从不自动提交、始终由用户做最终决定。将作者声称的 740+ 职位评估与 100+ 定制简历标注为\"作者声称\"而非独立指标。章节结构为：分类工位轨道 → 门户接入 → 评分档位 → 人工闸杆 → 批量并行轨道 → 快速开始 → 结果与局限。","en":"Use the actual pipeline language: scan job portals, evaluate with A-F scored blocks into a 1.0-5.0 score, keep the Block G posting-legitimacy check, generate tailored ATS CVs, and track applications; state plainly that the system never submits applications and the user always has the final call. Frame the author's claimed 740+ offers evaluated and 100+ tailored CVs as an author claim, not an independent metric. Section structure: triage rail → portal inbound → grade notches → manual lever → batch lane → quickstart → outcomes & caveats."},"forbiddenPatterns":["three metric cards under a slogan","terminal or CLI shell chrome as the page shell","browser or editor shell as the visual container","decorative glow, particle or grid backdrop used as a substitute for a project-specific visual idea","fake product screenshot, dashboard UI, or fake controls (tabs/search/switches)","generic feature-card wall that could accept any project's copy"],"historicalReuse":{"sourceAssetId":"cmsv8ronf0075bnqt6iuppfes","decision":"rebuild","rationale":"The historical asset is a dark 'command center' terminal dashboard (gradient glow, CLI shell chrome, panel glass). Its shell grammar is a deprecated default and cannot carry the V4 evidence. The only things worth saving are compositional ideas that still fit current facts: the single operational-surface layout (one surface, sections as panels), the mono/typewriter station-readout typography, and the stamped-tick quality of its status dots. Those are rebuilt onto the selected triage-rail mechanism rather than the old shell. Reusing the old HTML directly as V4 is rejected.","reusedElements":["single-unchanging operational surface with section panels (ideas only, rebuilt as machined rail, not glass cards)","monospaced/typewriter station-readout typography (rebuilt for rail labels)","status dot/claim-stamp typographic treatment (rebuilt as rail annotations and author-claim stamps)"],"discardedElements":["terminal/CLI shell chrome and dark command-center frame","gradient and radial glow background","glass panel cards and blur surfaces","any code, copy, unsupported numbers, or claim wording from the old HTML"]}},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"approved","summary":"Triage-rail design faithfully and project-specifically implements direction-b. The derived signature visual (horizontal station rail INCOMING->GRADE->TAILOR->TRACK with a red over-engineered G safety lever and decision notch, rebuilt vertical on mobile) is genuinely present in both locales. All claims are evidence-bound: the four pipeline stations, Block G legitimacy gate, 10+ parallel sub-agent batch as an annotation, and the author's 740+/100+ figures explicitly stamped as '作者声称/Author claim'. Quickstart shows only authentic readme commands. No browser/terminal shell chrome, no generic hero/three-card/card-wall, no decorative gradient-grid crutch, no fake controls. Both locales are structurally equivalent with natural, terminology-faithful copy.","checks":{"projectFit":{"passed":true,"evidence":"Both pages implement the exact selected direction-b sectionSequence (triage-rail, portal-inbound, grade-notches, manual-lever, batch-lane, quickstart, outcomes-caveat) in order. Visual language is a machined control-panel surface (plate #171c22, raised #222a34, recessed rails, stamped ticks, signal colors) matching surfaceModel/palette; condensed grotesque labels + mono station readouts match typography guidance."},"evidenceBound":{"passed":true,"evidence":"Four stations INCOMING/GRADE/TAILOR/TRACK derive from the verified pipeline language in projection.description. Block G legitimacy check is rendered as the separate break-lever per summary claim. '子代理/10+ 并行/批量评估' batch annotation is sourced from the parallel sub-agents claim and stays a labeled annotation, not an invented metric. 'remotli/getro' provider tokens come from release v1.26.0 evidence. 740+ offers / 100+ CVs / Head of Applied AI are explicitly framed as author claim ('作者声称'/'Author claim'), matching the caveat evidence."},"signatureVisual":{"passed":true,"evidence":"Real project-specific visual present in both locales: an SVG horizontal folded timeline rail with offer tokens entering left (x=286,600), a central grade dial (1.0-5.0), four station blocks, and a red over-engineered G lever pivoting toward a green decision notch at x=1010. Not a generic hero, not a screenshot, not an abstract gradient."},"mobileTransformation":{"passed":true,"evidence":"A genuine restructuring (not just scaling): @media(max-width:820px) hides .rail-desktop and reveals .rail-mobile, a stacked vertical flow where each station is a full row (vstamp) with per-row captions and a big centered G toggle (vlever) at the foot, matching the brief's 90-degree rotation instruction. QA mobile metrics confirm no horizontal overflow (scrollWidth 390 / clientWidth 390)."},"localeParity":{"passed":true,"evidence":"zh-CN and en share identical section structure, station labels, command set, the G-gate break lever annotation and the author-claim stamp. Chinese copy uses faithful pipeline terms (扫描职位门户/A-F 评分块折算为 1.0-5.0/定制简历/跟踪申请/Block G 合法性检查/系统从不自动提交) with natural language, English uses equivalent native phrasing. Both keep 'the user always has the final call'. Both frame the numbers as an author claim."},"antiTemplate":{"passed":true,"evidence":"No browser/editor/terminal shell wraps the page; the rail mechanism is the container. No generic three-metric-cards-under-slogan, no generic feature-card wall, no unchanged historical 'command center' dashboard chrome (historicalReuse=rebuild and discarded). The faint body radial gradients (<6% opacity) are ambiance only and do not substitute for a project visual, since the rail and lever supply the derived idea. Quickstart is presented as authentic command tags/cmdline text, not a feature grid."},"noFakeUi":{"passed":true,"evidence":"All surfaced 'controls' are diagrammatic or copy-only: the rail is an svg role=img aria-hidden visualization, the mobile lever is an inert visual inside a role=img block, cmdline blocks are styled text of real readme commands with # comments. No tabs, switches, buttons, search boxes, or interactive fake UI. QA reports fakeControlCount=0, brokenImageCount=0 across all four viewport/locale runs."}},"issues":[]},"repairEvidence":[{"attempt":1,"before":{"qaPassed":false,"designReviewApproved":false,"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":false,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":7,"textLength":1556,"cjkCount":766,"latinLetterCount":418,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":3354,"h1Top":166.21875,"bottomInset":207.984375,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"ca3fbdb6b3098a6ce34d77f40b7803a51abfb617143b81d77ba6526be43a0532","middle":"68a4d5d030d72c4eff4cef01d35386758d669314205b91915ba345f730b03280","bottom":"541c6528b07303ee3de839b4b43e1e96eb8001a37f147ba6be59be301bc1c32c"},"consoleErrors":["Error:

$\"career-ops\"$

[English](README.md) | [Español](README.es.md) | [Deutsch](README.de.md) | [Français](README.fr.md) | [Português (Brasil)](README.pt-BR.md) | [한국어](README.ko-KR.md) | [日本語](README.ja.md) | [简体中文](README.cn.md) | [繁體中文](README.zh-TW.md) | [Українська](README.ua.md) | [Русский](README."}]},"userIntents":[{"slug":"build_software","label":"构建或改进软件","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)"}]},{"slug":"automate_work","label":"自动化工作流程","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)"}]},{"slug":"create_content","label":"创作内容或设计","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)"}]},{"slug":"analyze_data","label":"分析与理解数据","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)"}]},{"slug":"learn_explore","label":"学习与探索","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)"}]}],"readme":{"headings":["What Is This","The CareerOps Manifesto","Features","Quick Start","2. Check setup","3. Configure","4. Add your CV","Create cv.md in the project root with your CV in markdown","5. Open your AI CLI in this directory","Then ask your CLI to adapt the system to you:","\"Change the archetypes to backend engineering roles\"","\"Translate the modes to English\"","\"Add these 5 companies to portals.yml\"","\"Update my profile with this CV I'm pasting\"","6. Start using","Paste a job URL or JD text to trigger auto-pipeline","If your CLI supports slash commands, use /career-ops (or its CLI-specific alias)","In Codex, ask for the same mode in plain language, e.g.:","\"Run the career-ops scan mode\"","\"Run the career-ops pipeline mode for data/pipeline.md\""],"commands":["npx @santifer/career-ops init","npx playwright install chromium # only needed for PDF generation","npm run doctor # Validates all prerequisites","npm i -g @santifer/career-ops","npm install","npm run gemini:eval -- \"JD text here\"","npm run serve:dashboard # launch the TUI","npm run build:dashboard # optional: build the standalone binary"]},"changeSet":[{"field":"authorName","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"967a44a3cc4c88f0187242f6fb30671134ea273f22d7a8774a388b258344a016"},{"field":"description","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"7e70999a1cecebf1a85924dc9b511cb10ea5c4b82474addcc5de9c90fb1ac5aa"},{"field":"experienceUrl","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"733d1608dbe986cabbf67a2a53dcaced2e77d28a2f126ca92d1825ec6bd1bf0a"},{"field":"name","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"6dd72df1a9aef015150a0f2b561065d9165bcd17029639ff5c5230d044977980"},{"field":"repository","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"00207dc81223593bffc7324c57356dbaf075a5fce743ac354708192fc2fa936a"},{"field":"sourceCount","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"6b86b273ff34fce19d6b804eff5a3f5747ada4eaa22f1d49c01e52ddb7875b4b"}]},"publishedAt":"2026-08-17T14:12:19.835Z","updatedAt":"2026-08-21T13:26:53.509Z","summary":{"id":"cmt2zh2qz01hv627102pshxhr","projectType":"product","oneLine":"将 AI 编码 CLI 变成求职指挥中心","overview":"career-ops 是一个开源 AI 求职系统，运行在 Claude Code、Codex 等 AI 编码 CLI 中。它扫描职位门户，用 A-F 评分块和 1.0-5.0 分数评估职位，生成 ATS 优化的定制简历，并跟踪申请流程。系统强调人机协作，从不自动提交申请，适合希望用 AI 辅助筛选职位、准备面试和优化求职流程的求职者。","highlights":["A-F 评分评估","批量职位扫描","定制简历生成","面试准备套件"],"bestFor":["求职者","AI 编码 CLI 用户","希望自动化求职流程的人"],"caveats":["首次评估效果不佳，需提供个人背景","需要配置 AI 编码 CLI 环境","作者声称的效果未经独立验证"],"categorySuggestion":"productivity","subcategorySuggestion":"automation_workflows","lifecycleStages":["idea_requirements","product_design","testing_quality"],"audienceIntents":["evaluate","build","integrate","operate"],"confidence":0.85,"evidence":[{"claim":"career-ops 将任何 AI 编码 CLI 转变为求职指挥中心，可扫描职位门户、评估职位列表并生成定制简历。","sourceType":"readme","locator":"https://github.com/santifer/career-ops"},{"claim":"使用 A-F 评分块和 1.0-5.0 分数评估职位，并包含职位合法性检查（Block G）。","sourceType":"readme","locator":"https://github.com/santifer/career-ops"},{"claim":"支持批量处理，通过子代理并行评估 10+ 个职位。","sourceType":"readme","locator":"https://github.com/santifer/career-ops"},{"claim":"系统从不自动提交申请，始终由用户做最终决定。","sourceType":"readme","locator":"https://github.com/santifer/career-ops"},{"claim":"提供面试准备、谈判脚本、后续跟进和模式分析等功能。","sourceType":"readme","locator":"https://github.com/santifer/career-ops"},{"claim":"作者声称使用该系统评估了 740+ 职位，生成了 100+ 定制简历，并获得了 Head of Applied AI 职位。","sourceType":"readme","locator":"https://github.com/santifer/career-ops"},{"claim":"最新版本 v1.26.0 增加了新的职位板提供商（如 remotli、getro）和封面信签名块等功能。","sourceType":"release","locator":"https://github.com/santifer/career-ops/releases/tag/career-ops-v1.26.0"}],"publishedAt":"2026-08-17T14:12:19.835Z"},"github":{"url":"https://github.com/santifer/career-ops","ownerAvatarUrl":"https://avatars.githubusercontent.com/u/256850418?v=4","starsCount":64288,"forksCount":12640,"language":"JavaScript","syncedAt":"2026-08-17T14:10:07.451Z","license":"MIT","defaultBranch":"main","archived":false,"disabled":false,"lastChangedAt":"2026-08-17T02:32:02.000Z","missing":false},"tags":[],"updates":[]},"entryPoint":"discovery","claim":{"state":"unclaimed","verified":false},"initialView":"beautified","community":{"viewer":null,"state":{"likesCount":0,"creativeCount":0,"followersCount":0,"commentsCount":0,"liked":false,"creative":false,"followed":false},"comments":{"comments":[],"nextCursor":null}}}]}],"$L24"]

把每一个职位，放到一条人工把关的分流轨道上