菜单
Batch parallel evaluation of multiple job offers via subagents. Triggers on batch evaluate, batch process, mass evaluate, parallel jobs.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Full career operations suite — evaluate job offers (A-G scoring), compare multiple offers, auto-pipeline from URLs, portal scanning, batch processing, ATS-optimized PDF/LaTeX CV generation, interview prep, application assistance, LinkedIn outreach, deep company research, application tracking, rejection pattern analysis, follow-up cadence, training evaluation, and portfolio project evaluation. Triggers on job, career, offer, oferta, bewerbung, CV, resume, interview, scan, pipeline, tracker.
Full A-G evaluation of a single job offer. Scoring across 6 dimensions (match, North Star, comp, culture, red flags, global) plus posting legitimacy analysis. Triggers on evaluate, oferta, review job, score job, assess offer.
Live application assistant — reads the form on screen, matches to evaluation context, generates tailored responses for each question. Triggers on apply, fill form, application form, apply to, bewerben.
Full automatic pipeline — paste a JD URL or text and get evaluation, report, PDF, draft answers, and tracker entry. Triggers on auto-pipeline, full pipeline, evaluate and apply, end to end.
Compare multiple job offers side-by-side with a weighted scoring matrix across 10 dimensions. Triggers on compare offers, compare jobs, ofertas, multiple offers, side by side.
Generate LinkedIn outreach messages tailored by contact type (recruiter, hiring manager, peer, interviewer). Triggers on linkedin, outreach, contact message, contacto, reach out.
| name | batch |
| description | Batch parallel evaluation of multiple job offers via subagents. Triggers on batch evaluate, batch process, mass evaluate, parallel jobs. |
Read ~/.openclaw/workspace/skills/career-ops/references/scoring-system.md first.
~/.openclaw/workspace/career-ops-data/batch/batch-input.tsv~/.openclaw/workspace/career-ops-data/batch/batch-state.tsv~/.openclaw/workspace/career-ops-data/reports/~/.openclaw/workspace/career-ops-data/data/applications.mdEach job is processed by a subagent (spawned via OpenClaw sessions_spawn) with a clean context. The main agent orchestrates.
batch/batch-state.tsv to identify what's already processedbatch-state.tsv (completed/failed + score + report_num)
d. Log to batch/logs/{report_num}-{id}.logbatch/tracker-additions/ → applications.mdEach subagent receives the evaluate skill prompt plus:
| Error | Recovery |
|---|---|
| URL inaccessible | Mark failed, continue |
| JD behind login | Mark failed, continue |
| PDF fails | Report saved, PDF pending |
| Subagent timeout | Mark failed, can retry with --retry-failed |
node ~/.openclaw/workspace/skills/career-ops/scripts/merge-tracker.mjs — merge TSV additions into trackernode ~/.openclaw/workspace/skills/career-ops/scripts/dedup-tracker.mjs — deduplicate tracker entries