aiden0602/cv-tailor

CV Tailor

Use this skill when the user asks to tailor a CV for a specific role, using:

• Job posting URL
• Copied JD text
• Screenshot/photo of JD text (OCR/vision-capable flow)

Inputs accepted from Telegram

1. Job URL
2. JD pasted text
3. JD image/capture

If image text cannot be read reliably, ask user for pasted text fallback.

Built-in resources

• file_skill.ts: safe read/write/list helper rooted at ~/openclaw_pro
• prompts/ATS_ALIGNMENT_ENGINE.md
• prompts/IMPACT_REFRAMING_ENGINE.md
• scripts/export_bullets_to_drive.sh
• scripts/cv_model_mode.sh

Standard runbook

0. Enter CV high-quality model mode (required):
   • bash scripts/cv_model_mode.sh start
1. Parse JD from URL/text/image.
2. Run ATS alignment analysis (keyword gaps, bullet rewrites, skills list).
3. Run impact reframing (summary + before/after impact bullets).
4. Read master_cv.md and tailor truthfully.
5. Save full version to cv/CV_[Company]_[Role].md.
6. Produce bullet-only version for copy/paste.
7. Upload bullet file to Google Drive (optional):
   • bash scripts/export_bullets_to_drive.sh "$HOME/openclaw_pro/cv/CV_[Company]_[Role].md"
8. Report saved local path + Drive file link/id.
9. Exit CV high-quality model mode (required, always run even on failure):
   • bash scripts/cv_model_mode.sh end

Safety constraints

• Never fabricate experience, projects, skills, or credentials.
• Keep claims grounded in user-provided CV/history only.
• For unknown metrics, request estimate or use qualitative impact.

Quality mode (CV-only)

• CV tasks must use google/gemini-2.5-pro while the task is running.
• Normal tasks should remain on stable default (google/gemini-2.5-flash).
• If provider outage/rate-limit forces fallback, explicitly report fallback in response.