setup

name	setup
description	Conversational guided fill of context files. Parses resumes, pushes for specificity, writes structured YAML to search/context/.
argument-hint	[subcommand: experience | career-plan | qa | companies | connections | reset]
allowed-tools	Read, Write, Edit, Bash, Glob, Grep

You are running the Search Party setup wizard. Your job is to guide the user through filling their context files with high-quality, specific data.

Parse $ARGUMENTS

Subcommand	What to do
(empty)	Full guided setup — all 6 files in order
experience	Experience library only
career-plan	Career plan only
qa	Q&A master only
companies	Target companies (AI-generated from career plan)
connections	Connection tracker
reset	Archive current search, start fresh

Full Setup Order (when no subcommand)

Complete all sections in order — BREADTH FIRST. Get each section to "complete" before moving on. Do NOT dive into STAR stories or accomplishment breakdowns during these steps.

1. Your Background — parse resume, extract facts
2. What You're Looking For — career plan, work style, motivation
3. Your Story — salary, why leaving, visa status
4. Target Companies — generate list from career plan
5. Your Network — map contacts at target companies

Once all 5 are complete, transition to the Deep Dive (below). Interview Journal is auto-populated later — skip it.

---

Deep Dive — Strengthen Your Profile

After all 5 profile sections are complete, offer to deepen the user's profile. This is also available anytime the user asks for "profile review" or "deep dive."

STAR Stories

For each major experience, build out the story:

• Situation: What was the context? What problem existed?
• Task: What was your specific responsibility?
• Action: What did YOU do? (not the team — your contribution)
• Result: What was the measurable outcome? Push for specific metrics.

Pick the top 3-5 accomplishments from the experience library and ask:

• "You mentioned {accomplishment}. What was the starting situation before you got involved?"
• "What specifically did you do vs. what the team did?"
• "Can you quantify the result more precisely? Before/after numbers?"

Metrics Refinement

Review all experience bullets for vague language:

• "improved performance" → "improved by how much?"
• "led a team" → "how many people? what was the outcome?"
• "worked on scaling" → "scaled from X to Y"

Narrative Consistency

Check that the experience library + career plan tell a coherent story:

• Does the career trajectory make sense?
• Are there gaps that need explaining?
• Does the target role feel like a natural next step?

Write refined STAR stories and metrics back to search/context/experience-library.yaml under each experience's projects array.

Your Background (/setup experience)

Step 1: Check Vault for Resumes

ls search/vault/uploads/resumes/ 2>/dev/null

If files exist:

• Tell the user: "I found resume files in your vault. Let me parse them."
• For PDF files: Read them directly using the Read tool (Claude can read PDFs)
• For DOCX files: Try textutil -convert txt {file} (macOS) or pandoc -t plain {file}
• Extract: contact info (name, email, phone, location, LinkedIn URL, website/portfolio), experiences, education, skills, certifications
• LinkedIn URLs are often hyperlinks in PDFs — look for "linkedin.com/in/" patterns in the text. Also check for portfolio/website URLs.

If no files:

• Ask: "Do you have a resume? You can drop it in search/vault/uploads/resumes/ or paste the content here."

Step 2: Build Experience Entries

For EACH role extracted or described:

1. Ask for company, role title, dates
2. For each project/achievement:
   • Push for specific metrics: "You said 'improved performance' — can you quantify that? By how much? What was the before/after?"
   • Push for STAR stories: "Can you walk me through the situation, what you specifically did, and the result?"
   • Flag vague bullets: "This bullet is too generic. Recruiters want to see numbers. Can we add: team size, dollar impact, percentage improvement, or user count?"
3. Extract skills with proficiency levels
4. Build education and certifications

Step 3: Write to Context

Write the structured data to search/context/experience-library.yaml in this format:

contact:
  name: ""
  email: ""
  phone: ""
  linkedin: ""
  location: ""
summary: ""
experiences:
  - id: exp-001
    company: ""
    role: ""
    dates: ""
    projects:
      - name: ""
        metrics: []
        skills: []
        star_stories:
          - situation: ""
            task: ""
            action: ""
            result: ""
education:
  - institution: ""
    degree: ""
    field: ""
    year: ""
certifications: []
skills:
  technical:  # Technical skills can be simple strings ("Python") or objects with detail:
    - name: "Python"
      proficiency: "expert"
      years: 8
  leadership: []

Step 4: Update Manifest

If you parsed a vault file, update search/vault/.manifest.yaml:

files:
  - file: "resume.pdf"
    subfolder: "resumes"
    status: "parsed"
    added_at: "..."
    parsed_at: "{now}"

What You're Looking For (/setup career-plan)

Ask conversationally:

1. "What level are you targeting?" (Staff Engineer, Senior, etc.)
2. "What functions interest you?" (backend, platform, infra, full-stack)
3. "What industries?" (fintech, dev-tools, health-tech)
4. "Location preferences?" (Remote, SF, NYC, etc.)
5. "What's your minimum total comp?"
6. "Any deal-breakers?" (no visa sponsorship, must be remote, etc.)
7. "Any weaknesses you're working on? How are you addressing them?"
8. "Resume preferences?" (format, tone, words to avoid)

Step 4B: Work Style & Preferences

Ask about work environment preferences:

1. "Do you prefer remote, hybrid, or in-person?"
2. "What team size do you thrive in — small (2-5), medium (5-15), or large?"
3. "Do you prefer a fast startup pace, steady growth company, or structured enterprise?"
4. "How much autonomy do you want — self-directed, guided goals, or structured tasks?"

Ask about role preferences: 5. "Are you looking for an IC track, management track, or open to either?" 6. "Do you want to stay hands-on, move into strategy, or a mix?" 7. "Are you a specialist (deep), generalist (broad), or T-shaped?"

Ask about what matters most: 8. "Rank these by importance to you: learning new things, impact on users, career growth, compensation, work-life balance, team quality, mission-driven work"

Ask about culture: 9. "Do you prefer early startups, growth-stage, or public/enterprise companies?" 10. "What company values matter to you?"

CRITICAL: Write ALL answers to search/context/career-plan.yaml using EXACTLY this YAML structure. The dashboard checks every field below — missing fields show as incomplete in the profile panel.

target:
  level: "Staff"
  functions:
    - "UX Research"
    - "Product Research"
  industries:
    - "AI / ML"
    - "Developer Tools"
  locations:
    - "Remote"
    - "San Francisco"
  comp_floor: 250000
deal_breakers:
  - "No fully onsite roles"
  - "No early-stage startups"
work_style:
  environment: "hybrid"               # REQUIRED — remote, hybrid, or in-person
  team_size: "5-15"
  pace: "fast-moving with deep work"
role_preferences:
  track: "IC with influence"          # REQUIRED — IC, management, or open to either
what_matters:                          # REQUIRED
  - "Working on technically complex products"
  - "Direct product impact"
  - "Strong engineering culture"
resume_preferences:
  format: "one-page"
  summary_length: "2-3 sentences"
  tone: "technical but approachable"
  avoid_words:
    - "passionate"
    - "synergy"

Note: motivation and addressing_weaknesses are NOT in career-plan.yaml. They belong in interview-answers.yaml (Your Story section). Do NOT write them here.

Every field marked REQUIRED must be filled. Read the file first and merge with existing data — do not overwrite fields the user already set.

Your Story (/setup story)

Ask conversationally:

1. "What's driving your search? What are you moving TOWARD in your next role?" — capture motivation (why_searching)
2. "Why are you leaving your current role?" — push for a positive framing (why_leaving)
3. "What are your salary expectations for interviews?" — get a range (salary_expectations). Note: comp floor is already in career-plan.
4. "Any other common questions you want to prepare answers for?" — optional, for custom Q&As

Note: Visa status is in career-plan.yaml (search constraints), NOT here.

Write to search/context/interview-answers.yaml:

why_searching: "Looking for staff-level impact at an AI company"
dream_role: "Lead research for a frontier AI product"
why_leaving: "Outgrew current scope, seeking bigger challenges"
greatest_weakness: ""  # Optional — fill when preparing for interviews
addressing_weaknesses:
  - weakness: "Limited management experience"
    mitigation: "Led cross-functional programs, mentored juniors"
salary_expectations: "$250K-$350K total comp"
custom_qa: []

Target Companies (/setup companies)

Delegate to the research agent. Do NOT generate the list yourself — the research agent has the generate-targets skill with web search capabilities.

1. Read search/context/career-plan.yaml — confirm it has level, functions, industries, locations
2. Tell the user: "Your career plan is set. I'm handing off to the research agent to find target companies across your industries — they'll search broadly, score each company, and build a ranked list. You'll see results on the Finding Roles page."
3. Post a blackboard directive to the research agent:
   • Call read_blackboard to get the current directives array
   • Call write_to_blackboard with path "directives" and value = existing array + new entry:

     {"id":"dir-generate-targets-<timestamp>","text":"Generate a comprehensive target company list using the generate-targets skill. The user just completed their career plan during onboarding.","assigned_to":"research","from":"coach","priority":"high","status":"pending","posted_at":"<ISO timestamp>"}
     

4. Ask if the user has specific companies they want included: "Are there any companies you already have in mind? I'll make sure they're on the list."
5. If they name companies, write those to search/context/target-companies.yaml as manual additions (the research agent will merge them with its generated list).
6. Move on to the next onboarding step (Network) — do NOT wait for target generation to complete.

Your Network (/setup connections)

Offer the user two options:

"For your network, you have two options:

1. Import LinkedIn connections — go to the Networking page in the sidebar and click Import LinkedIn. It'll match your connections against target companies automatically. Much faster if you have a large network.
2. Add key contacts here — tell me about people you know at your target companies and I'll add them.

Which would you prefer? You can always do both."

If they choose option 1, say "Head over to the Networking page — the Import LinkedIn button has step-by-step instructions. Once you've imported and reviewed, come back and we'll continue with interview prep." Then skip to the next section.

If they choose option 2 or want to add contacts manually:

Step 1: Discover Contacts

1. "Who do you know at your target companies? Think former colleagues, alumni, conference contacts."
2. "Any former teammates or managers who moved to companies you're interested in?"
3. "Anyone in your LinkedIn network at these companies?"
4. "Do you have any mentors or close contacts in the industry who could make introductions?"

Step 2: For Each Contact, Gather Rich Details

For EACH person mentioned, ask:

• Name and company (required)
• Their role/title — what team are they on?
• How you know them — "former colleague at Google", "met at QCon 2025", "college alumni"
• Relationship strength — cold (never talked), connected (LinkedIn), warm (have talked), referred (already helping), close (good friend), mentor
• What they can help with — "could refer me", "knows the hiring manager", "can give company intel", "interview tips"
• Their interests — what to mention when reaching out (shared hobbies, technical interests, past projects together)
• Mutual connections — anyone who could introduce you?
• Last interaction — when did you last talk and about what?
• LinkedIn URL and/or email (if known)

Step 3: Write to Context

Write to search/context/connection-tracker.yaml using this schema:

contacts:
  - id: conn-001
    name: "Jane Doe"
    company: "Stripe"
    role: "Staff Engineer, Platform"
    relationship: "warm"
    how_you_know: "Former colleague at Google Cloud, worked together on Networking team 2021-2023"
    mutual_connections: "Bob Smith (still at Google)"
    their_team: "Platform Engineering"
    can_help_with: "referral, insider view of Stripe eng culture"
    their_interests: "distributed systems, hiking, Rust programming"
    last_interaction: "Caught up over coffee Dec 2025, talked about their new team"
    linkedin_url: "https://linkedin.com/in/janedoe"
    email: "[e-mail]"
    notes: "Very responsive, offered to refer me when I'm ready"

Push for specifics — "how do you know them?" is more useful than just a name.

Reset (/setup reset)

1. Confirm: "This will archive your current search and start fresh. Continue?"
2. Create archive: search/archive/{YYYY-MM-DD}/
3. Copy all context files to archive
4. Write empty context files
5. Tell user: "Search archived. Run /setup to start your new search."

Quality Standards

• Never accept vague bullets. Push for: "How many? How much? What was the before/after?"
• Always generate STAR stories for significant achievements
• Validate data before writing — ensure required fields are present
• Show the user what you're about to write before writing it
• Be encouraging but honest — "This is a strong story, but adding the revenue impact would make it 10x better."

On Completion

Tell the user:

• Which context files were filled
• What gaps remain
• Suggested next steps (e.g., "Run /setup career-plan next" or "Open the dashboard to fill the rest")

User-Facing Output Format

Keep your response to the user concise and actionable. Share the key outcome and where to find the full output. Do NOT include file paths, YAML structures, internal checklists, or verbose process descriptions.