Purpose
Import an existing resume into resume.yaml format. Scans for existing documents, interviews the user to fill gaps, and writes the structured YAML. Assumes /setup has already run (project structure and CLAUDE.md docs exist).
Process
Phase 1 — Discovery
Scan the current directory and one level of subdirectories for existing resume artifacts. If the user provides an external path (e.g., /import ../resume/), scan that path as well.
Resume files: Scan for resume artifacts in any common format (YAML, markdown, PDF, DOCX, JSON, HTML). Prioritize resume.yaml if it already exists in resume-builder format.
External documents (when the user provides an external path):
- Read and understand any document format found (HTML exports, PDFs, markdown, text files, etc.)
- Skip asset/support directories that don't contain meaningful content
- Extract project names, technologies, architecture, metrics, and timelines
Report findings to the user:
Found in source directory:
✓ resume.md (8.8 KB) — markdown resume (will parse)
✓ projects/ (15 HTML exports) — project documentation (will parse)
✗ skills.yaml — NOT a plugin file (skill data goes in resume.yaml)
✗ generate_resume_pdf.py — NOT needed (replaced by /generate)
✗ index.html — NOT needed (replaced by /generate HTML output)
Phase 2 — Parse Existing Resume
If an existing resume file is found:
-
Read and parse it — extract all structured data:
- Header: name, title, location, email, phone (optional), LinkedIn, GitHub
- Summary paragraph and highlight bullets
- Skills by category
- Experience: companies, roles, dates, bullet points
- Projects: name, URL, description, technologies
- Education
-
Extract technologies and metrics from each bullet:
- Technologies: tools, frameworks, languages, services mentioned
- Metrics: percentages, counts, scale figures, latency numbers
-
Present the parsed structure to the user for review:
Parsed from resume.md: Header: Wilson K. Choi | Senior Software Engineer | ... Summary: 3 bullets Skills: 6 categories Experience: 3 companies, 6 roles, 20 bullets Projects: 4 entries Education: B.S. Computer Science, Cal Poly Pomona Does this look complete? Anything missing or incorrect?
If external documents are found:
-
Read each document — extract meaningful content, skipping boilerplate and asset directories.
-
Categorize extracted content per document:
- Technologies and tech stack
- Architecture patterns and design decisions
- Scale/volume metrics
- Timelines and delivery dates
-
Present summary to the user:
Parsed from 15 project documents: ASR for IVR — UniMRCP, FreeSWITCH, Azure, 70-80k calls/month Call Recording Architecture — S3, Lambda, transcription pipeline SIP Trunking — Kamailio, FreeSWITCH, HA design ... Which of these projects should be included in your resume?
Phase 3 — Interview
Ask clarification questions to fill gaps. Do NOT proceed until these are resolved:
Always ask:
- "Is any content in this resume outdated or no longer accurate?"
- "Are there any recent projects or roles not captured here?"
- "Which types of roles are you targeting?"
Ask if missing:
- Contact info gaps (email, phone, LinkedIn, GitHub)
- Date gaps in employment history
- Bullet points with no quantified impact — ask if metrics are available
- Projects with no URLs — ask if they're public
Do NOT ask:
- Questions answerable from the documents already scanned
- Obvious formatting preferences (the plugin handles this)
- Technical setup questions (the plugin handles this)
- Whether to keep legacy generation scripts (they're replaced by /generate)
Phase 4 — Confirm Before Writing
Present the complete plan before writing:
Ready to import. Here's what I'll write:
WRITE resume.yaml — 20 bullets, 4 projects, all provenance set to manual+verified
Provenance: all bullets marked source:"manual", verified:true
Proceed? [describe any changes you want first]
Wait for explicit user confirmation before writing.
Phase 5 — Write and Verify
After user confirms:
-
Write
resume.yamlwith full provenance on every bullet:provenance: { source: "manual", artifacts: [], verified: true } -
Run
resume-builder verifyto validate the file -
Report results:
✓ resume.yaml written — 20 bullets, 4 projects, all verified ✓ resume-builder verify passed Next steps: /generate — generate all 4 formats (PDF, DOCX, HTML, Markdown) HTML output is a portfolio-quality page with dark/light theme, spotlight effects, animations, and responsive design /ingest — add achievements from project artifacts in artifacts/ /tailor — tailor for a specific job description
Resume.yaml Schema Reference
The generated resume.yaml must conform to this structure:
header:
name: str
title: str
location: str
email: str
phone: str | None # optional
linkedin: str
github: str
summary:
paragraph: str
bullets:
- label: str
text: str
provenance: { source: str, artifacts: [], verified: bool }
skills:
- category: str
items: str # comma-separated
provenance: { source: str, artifacts: [], verified: bool }
experience:
- company: str
location: str
description: str (optional)
dates: str
roles:
- title: str
dates: str
description: str (optional)
bullets:
- label: str
text: str
technologies: [str]
metrics: [str]
provenance: { source: str, artifacts: [], verified: bool }
projects:
- name: str
url: str
description: str
technologies: [str]
provenance: { source: str, artifacts: [], verified: bool }
education:
degree: str
institution: str
Anti-Fabrication Rules (MANDATORY)
- NEVER fabricate or embellish content during import — copy what exists
- NEVER skip the interview phase — always ask the user to confirm before writing
- If the existing resume has content that looks wrong or inconsistent, flag it rather than silently copying
- Set ALL provenance to
source: "manual", verified: truefor content imported from user's existing resume - If
resume.yamlalready has content, ask before overwriting