career-intelligence-onboarding

name	career-intelligence-onboarding
description	Week 1 onboarding wizard for new Career Intelligence users. Interviews the user across 3 areas (work history, skills, job search preferences), then generates the 2 required context files that Career Intelligence needs to run. The outreach-composer, outreach-fact-check, resume-engine, and job-match-scorer skills all load these files automatically — the user never re-explains their background.
triggers	["onboard me to career intelligence","career setup","career onboarding","set up my career","career intelligence setup","set up job search","i'm new to career intelligence","get me started with career intelligence","career intelligence week 1"]

Career Intelligence Onboarding — Week 1 Setup

Purpose

New Career Intelligence users need 2 files before job search, outreach, and resume skills can run accurately. This skill generates both through a structured interview. Once complete, the user's experience history and job search preferences persist across every future session — no re-explaining background on every outreach draft.

Output Format

Always start with:

━━━ Career Intelligence: Onboarding ━━━

The 2 Files This Skill Generates

All writes go through brain.write() with engine_id: "career-intelligence". identity/experience-history.md is an xOS primitive — permitted via writes_to_primitives declaration.

File	brain.write() path	What it does
experience-history.md	identity/experience-history.md	Work history, key achievements, skills, education. Loaded by outreach-fact-check before every outreach draft to prevent fabricated claims.
job-search-config.md	career-intelligence/projects/job-search-config.md	Target roles, companies, location, salary, and non-negotiables. Loaded by job-match-scorer and mission-control.

Execution Flow

Phase 0 — Workspace Setup (runs FIRST)

Check $CAREER_HOME. If not set or directory doesn't exist:

"Where should your Career Intelligence workspace live? This is where your experience history, job pipeline, and session memory will be stored.

Default: ~/career-os (press Enter to accept, or type a different path)"

After they answer:

1. Create: mkdir -p "$CAREER_HOME/identity" "$CAREER_HOME/network/people" "$CAREER_HOME/career-intelligence"
2. Persist: echo 'export CAREER_HOME="<path>"' >> ~/.zshrc && export CAREER_HOME="<path>"
3. Print: "Workspace ready at $CAREER_HOME. Let's capture your background."
4. Emit the local-only onboarding start signal. The helper no-ops unless XOS_98_TELEMETRY is enabled:

   bun "$CLAUDE_PLUGIN_ROOT/src/telemetry/beta-funnel.ts" onboarding-started '{}'
   

If already set: confirm directory is writable, emit the same onboarding start signal, and proceed.

---

Phase 1 — Work History (4 questions, ONE AT A TIME)

Q1: "What is your current or most recent job title and company? (e.g., 'Senior Product Manager at Stripe')"

Q2: "Walk me through your last 3 roles — for each: title, company, and the ONE achievement you're most proud of. (Don't worry about perfect wording — rough is fine, we'll refine it.)"

Q3: "What are your top 5–7 technical or functional skills? (e.g., 'Python, product strategy, LLM systems, stakeholder management')"

Q4: "Highest education: degree, field, institution, year? Any other credentials worth noting (certifications, notable programs, executive education)?"

---

Phase 2 — Job Search Preferences (4 questions, ONE AT A TIME)

Q5: "What type of role are you targeting next? Give me: title/level, function, and type of company (startup, growth, enterprise, specific industry)."

Q6: "Location and work style preference? (e.g., 'SF Bay Area, hybrid 2x/week' or 'remote-first, any US timezone')"

Q7: "Target compensation range? Total comp including base + equity if relevant. (This stays local — used only for job-match scoring, never shared.)"

Q8: "What are your hard NON-NEGOTIABLES? Things that would make you immediately reject a role. (e.g., 'no contract roles, no defense/weapons, must have equity, visa sponsorship required')"

---

Phase 3 — File Generation

After all 8 answers:

1. Read the 2 template files:

   $(ls -v ~/.claude/plugins/cache/xos/career-intelligence/*/skills/career-intelligence-onboarding/onboarding-templates/*.template.md 2>/dev/null)
   

2. Fill in all {{PLACEHOLDER}} tokens with the user's answers. For multi-role work history (Q2): expand the template's role blocks to cover all roles mentioned.

3. Write the 2 files via brain.write():

   brain.write("identity/experience-history.md", content, {
     provenance: { who: "career-intelligence", why: "onboarding: experience history", source: "career-intelligence-onboarding" },
     engine_id: "career-intelligence"
   })
   brain.write("career-intelligence/projects/job-search-config.md", content, {
     provenance: { who: "career-intelligence", why: "onboarding: job search config", source: "career-intelligence-onboarding" },
     engine_id: "career-intelligence"
   })
   

   If files already exist (brain.exists() returns true): show what would change. Ask: "Overwrite, merge, or skip?"

4. Validate: confirm both files exist and are non-empty.

5. Smoke test: confirm experience-history.md has at least one role entry and job-search-config.md has a target_roles field.

6. Emit the local-only onboarding completion signal:

   bun "$CLAUDE_PLUGIN_ROOT/src/telemetry/beta-funnel.ts" onboarding-completed '{}'
   

---

Phase 4 — Completion Summary

Print this summary:

━━━ Career Intelligence Onboarding Complete ━━━

Files created:
  ✅ brain/identity/experience-history.md
  ✅ career-intelligence/projects/job-search/job-search-config.md

Cross-session memory: ACTIVE
  → Your background is now loaded into every outreach draft automatically.
  → outreach-fact-check will verify claims against your real history before sending.
  → job-match-scorer will score roles against your preferences — no re-explaining each time.

What's next:
  → "show me my job pipeline" — see your current applications
  → "score this job: [paste JD]" — instant fit analysis against your preferences
  → "write outreach to [name] at [company]" — grounded in your real background

One optional step: git commit your context files so they're versioned:
  git -C $CAREER_HOME add brain/identity/experience-history.md career-intelligence/projects/job-search/job-search-config.md
  git -C $CAREER_HOME commit -m "feat(career): Week 1 onboarding — experience history + job search config"

---

Failure Modes

• User skips a question: Ask again with context ("This becomes the ground truth for every outreach email — even a rough version catches fabrication errors before your emails go out.")
• User has many roles (>5): Ask them to prioritize the 3 most relevant to their target. Others can be added later by editing the file directly.
• User declines to share salary: Leave target_compensation as "not specified" — job-match-scorer skips compensation scoring.
• Files already exist: Show a diff preview. Ask: "Overwrite, merge, or skip?" Never silently overwrite.