Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add dizhouid-lgtm/liepin-ai-loop-recruiting --skill pxb-liepingit clone --depth 1 https://github.com/dizhouid-lgtm/liepin-ai-loop-recruitingWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/dizhouid-lgtm/liepin-ai-loop-recruiting/pxb-liepin)<a href="https://agentmods.dev/skills/dizhouid-lgtm/liepin-ai-loop-recruiting/pxb-liepin"><img src="https://agentmods.dev/badge/skills/dizhouid-lgtm/liepin-ai-loop-recruiting/pxb-liepin/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/dizhouid-lgtm/liepin-ai-loop-recruiting/pxb-liepin"><img src="https://agentmods.dev/badge/skills/dizhouid-lgtm/liepin-ai-loop-recruiting/pxb-liepin.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00152 | $0.09053 |
| Opus 5.5 | $0.00061 | $0.03621 |
| Sonnet 5.5 | $0.00030 | $0.01811 |
| Haiku 4.5 | $0.00015 | $0.00905 |
Grade A, and why
pxb-liepin scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Pixboom 搜人筛人作业流程
⚠️ 改本 Skill 前先想:这里只放所有岗通用、可复用的作业流程。单一案例/某轮/某岗经验 → 进该岗
JD.md,不进这里。 每岗四件:JD.md(对外 JD + 内部备注表,判断只看它)·去重台账.csv(resume_id,status,仅去重)·候选人池.md(确认候选人)·待定/(简历 PDF 中转)。 路径:命令里<岗位>\…相对工作区根(本CLAUDE.md所在);脚本全在.claude/skills/pxb-liepin/scripts/,命令都写全路径,原样可跑。机械活全在脚本、跨 Win/mac;判断在本文 + agent。
全流程:两层循环,主 agent 串行自跑
P. 聊全 JD → 0. 体检+抢锁 → 0.5 登录 → 1. 粗筛 → 2. 精筛 → 3. 够一批没(没够且没到批次上限回 1)→ 4. 交用户复核 → 5. 入池 → 6. 迭代回 0;招满/用户喊停 → 7. 结项复盘。
- 不派 subagent:单账号单锁串行,subagent 无并行收益、反而启动慢易卡死;主 agent 从头跑到尾。
- 小循环(1-3,自跑、不打扰用户):搜一页 → 粗筛挑 ~6 → 精筛 → 对口标
待定→ 没攒够回步骤 1,到「批次上限」为止(见下方风控纪律)。不要跑一轮就停下来等用户。 - 大循环(4-6,用户校准 JD):交一批 → 用户复核 → 入池 → 迭代回写 JD(重点是进化关键词)→ 下一批。JD 越调越准、小循环越搜越准。
- 上下文:脚本只输出窄表。看完做完判断后记住结论即可(
待定的 id=标签、方向感、哪个词有效),别把整页窄表一直留着;轮数多了丢旧轮明细。
状态机(去重台账只此 5 值,全程走 dedup.mjs,绝不手改 CSV)
未精筛(捞回来没细看的干净库存)→ 粗筛不合适(卡片就否)/精筛不合适(细看后否,含用户毙掉的)/待定(细看够格、出了 PDF 等复核)→ 入选(用户拍板要了)。
心跳 + 输出纪律(治"闷头静默"和"过程废话/散问打断")
- 边跑边报:每跑完一轮先发一行进度再继续,别让用户 10-20 分钟看不到动静。格式:
第N轮(词:X):搜A/精筛B/新增待定C,累计 D/目标 E,继续…。心跳是单向,不算打扰。 - 心跳就那一行:逐人淘汰理由、本轮发现、推理全留你脑子里 + JD 表,别讲给用户。
- 交付干净:出批消息 = 几份可点 PDF + 问为什么,别夹每轮过程、复盘、SKILL 自测。
- 决策别打断、给推荐、能不问就别问:破例/放宽约束/停搜这类才需用户拍的,优先带推荐先做可逆一步(照推荐先提报、复核可撤),决策点在交付里一句话点出;真要问就攒一处、一次问清、每条附推荐,别过程里东一句西一句、也别没问完就干等。照 JD 能自己判的,别问。
全程红线
① 单账号串行,绝不并发调 liepin。② 反爬宁慢勿封:脚本非零退出即停、别密集连刷(脚本内已带抖动)。③ 原始 JSON 不进上下文(脚本只输出窄表)。④ 隐私:简历只存本地、不外发、不传外部服务。⑤ 只 search/resume,不 greet(沟通归 HR)。⑥ 只走脚本(search/fetch/jd-round),除 liepin login 外不要直接敲 liepin 命令。search/fetch 强制有头(会弹 Chrome 窗口,属正常现象;脚本会忽略环境变量里的 LIEPIN_HEADLESS)。开搜前提醒用户:窗口别关,也别在里面操作。不要打 patch-headless.mjs(已废弃)。
风控纪律(优先级高于其他所有步骤)
实测结论(2026-10):触发风控的主要是浏览器冷启动、无头模式和关浏览器后重登,不是调用量。冷启动(
liepin login拉起新浏览器)当场就弹过验证。浏览器常驻有头时,一把锁内连跑 10 次 search、约 26 份简历加 5 份 PDF,共 2 小时以上,都没有触发。 猎聘风控有两种表现:页面被清成 about:blank;或被 302 到safe.liepin.com/.../captchaPage_PC(账号判为「行为异常」)。触发后所有业务接口都会被静默拦截,resume只回一个{"flag":0}(这不是权益不足)。按下面几条执行:
- 退出码 3 = 风控,立刻全停:当场释放锁,不再调任何 liepin 命令,包括"看一眼报错"式的重跑和
doctor --kill。告诉用户:「猎聘要求做一次安全验证,我打开窗口,你点『点击验证』完成滑块」,然后跑liepin login。滑块必须用户自己做,我不能代做。验证通过后先只跑一条 search 试探,通了再按节奏继续。 - 浏览器常驻,不要关:猎聘的登录 cookie 是会话级(
is_persistent=0),关掉浏览器就等于退出登录,反复"关浏览器→重登"正是被判「行为异常」的主因。所以:不要doctor --kill(除非浏览器确实卡死,而且用户同意);liepin quit只在当天收工时用;不要在有头和无头之间切换(切换就要关浏览器)。脚本超时强杀时也会清掉浏览器,所以超时之后按"已登出"处理,先确认登录态再继续。浏览器没在跑时(doctor 报"CDP 端口上没有浏览器"),先liepin login拉起来,并提前告诉用户:冷启动大概率会弹一次滑块验证,需要用户自己完成。doctor 报的"残留 liepin 浏览器进程"如果就是常驻浏览器的子进程(CDP 端口上有有头浏览器),那是正常现象,不要杀。 - 登录要省着用:
liepin login默认复用现有会话;24 小时内交互式登录超过 3 次会被 CLI 拦下,不要加--force硬闯。 - 批次上限:一把锁(一批)最多跑 10 次 search,每轮精筛最多 6 份,一批简历合计不超过 30 份(PDF 计入)。这是实测跑通的量级(见本节开头),不是 CLI 的限制;超出部分还没验证过。到上限还没凑够 5 个,手上有几个
待定就交几个(≥1 即交),同时报告召回情况;用户明确要求继续时可以开下一批。召回枯竭的信号是一次 search 的 40 人里新人不到 10 个,这时与其硬刷,不如先挖未精筛库存或建议放宽条件。 - 出 PDF 只用新版
pdf.mjs:它复用常驻的有头浏览器、新开一个标签页截图,结束只关自己的标签页,不关浏览器、不切无头。旧版(无头page.pdf(),要先关浏览器)两次风控前都跑过,已废弃。PDF 只给待定出,每份照样算一次看简历,计入批次上限的 30 份。
What ships with it
11 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- scripts/_csv.mjs 1.7 KB runs code
- scripts/_liepin.mjs 7.5 KB runs code
- scripts/dedup.mjs 2.7 KB runs code
- scripts/doctor.mjs 5.8 KB runs code
- scripts/fetch.mjs 3.1 KB runs code
- scripts/init-role.mjs 1.4 KB runs code
- scripts/jd-round.mjs 1.6 KB runs code
- scripts/lock.mjs 3.2 KB runs code
- scripts/patch-headless.mjs 2.0 KB runs code
- scripts/pdf.mjs 14 KB runs code
- scripts/search.mjs 4.0 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago Changed · +14 lines ff8439840e60
- 12d ago Changed 525f7eb99866
- 1mo ago First seen · 211 lines · 152 tokens per session scan A 06f5bd5f0b3f
pxb-liepin is a skill published in the GitHub repository dizhouid-lgtm/liepin-ai-loop-recruiting (3 stars, last pushed 11d ago), licensed MIT. It adds 152 tokens to every session and 9,053 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
customer-journey-map
Construye un mapa de recorrido del cliente para un producto, servicio o experiencia. Úsalo cuando se te pida mapear un recorrido del cliente, crear un viaje del usuario, documentar puntos de contacto y puntos de fricción, o diseñar un mapa de experiencia. Produce un mapa de recorrido completo con etapas, puntos de…
scenario-zbrush-expert
Use when an agent drives Maxon ZBrush 2026 for any task (sculpt in ZBrush, DynaMesh, ZRemesher, UVs, export, polypaint, posing, 3D print, clean up an AI mesh in ZBrush), when writing ZBrush Python (zbrush.commands) or ZScript, when the ZBrush bridge is not answering, a palette path or set() does nothing, a scripted…
feature-prioritisation
Aplica marcos de priorización (RICE, MoSCoW, Kano, ICE, Opportunity Scoring) para clasificar características e items del backlog. Úsalo cuando se te pida priorizar características, clasificar un backlog, decidir qué construir a continuación, o evaluar compensaciones entre ideas competidoras. Produce una lista de…
design-handoff-brief
Transform feature briefs into structured design briefs that give designers the context they need before opening Figma. Use when asked to write a design brief, create a design handoff, brief a designer on a new feature, or translate a PRD into design requirements. Produces a brief with user goal, emotional context…
outcome-tracker
Record the testable predictions inside a decision, then score them against reality later — so frameworks earn trust from outcomes, not vibes. Use when committing to a prioritisation, forecast, or plan (to log what it predicts), when asked to review what actually happened, or to compute how well-calibrated past RICE…
explorer-pipeline
Use when asked to make an interactive 3D explorer of a topic from start to finish, go from a topic to a clickable 3D learning page, or run the whole 3D explorer process. Produces a staged run of explorer-interface-brief, model-prompt-pack and explorer-viewer-builder in order for one topic, with the hand-offs between…