auto-pipeline

name	auto-pipeline
description	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.

Auto-Pipeline — Full Automatic Pipeline

Read ~/.openclaw/workspace/skills/career-ops/references/scoring-system.md first.

Data Paths

• CV: ~/.openclaw/workspace/career-ops-data/cv.md
• Profile: ~/.openclaw/workspace/career-ops-data/config/profile.yml
• Reports: ~/.openclaw/workspace/career-ops-data/reports/
• Tracker: ~/.openclaw/workspace/career-ops-data/data/applications.md
• Output: ~/.openclaw/workspace/career-ops-data/output/

Workflow

Step 0 — Extract JD

If URL: Use Playwright (preferred) → WebFetch (fallback) → WebSearch (last resort). If text: use directly.

Step 1 — A-G Evaluation

Execute full evaluation (see evaluate skill). Include Block G Legitimacy.

Step 2 — Save Report

Save full evaluation in reports/{###}-{company-slug}-{YYYY-MM-DD}.md. Include URL: and Legitimacy: in header.

Step 3 — Generate PDF

Read config/profile.yml. Check cv.output_format:

• If "latex", execute LaTeX pipeline (see latex skill)
• Otherwise, execute HTML/PDF pipeline (see pdf skill)

Run: node ~/.openclaw/workspace/skills/career-ops/scripts/generate-pdf.mjs

Step 4 — Draft Application Answers (only if score >= 4.5)

If score >= 4.5:

1. Extract form questions (Playwright or generic)
2. Generate responses following "I'm choosing you" tone
3. Save as section H in the report

Tone rules:

• Confident without arrogance
• Specific and concrete — reference something REAL from JD and something REAL from CV
• Direct, no fluff, 2-4 sentences per response
• The hook is the proof, not the statement

Step 5 — Update Tracker

Record in data/applications.md with all columns. If any step fails, continue and mark as pending.

German Mode

Read references/german-locale.md for DACH-specific checks.