Job Search Setup

Job Search Setup

The foundation. Builds workspace/profile/ and the master resume, the source of truth every other skill depends on. Nothing downstream works well until this runs.

When to use

• First run of the OS, or the orchestrator routes here because workspace/profile/ is empty or thin
• "Set me up", "onboard me", "build my profile", "start my job search", "update my targets"
• The user changes direction (new target titles, new comp floor, relocating) and profile/ needs a rewrite

Inputs

• Reads: an existing resume (pasted text, a file path, a PDF, or a LinkedIn export) if the user has one, plus workspace/profile/ and workspace/documents/ to check what already exists
• Needs from user: their resume (or permission to build one from scratch by interview), target titles, comp floor, location/remote, timeline, deal-breakers. Ask for the resume up front. Do not proceed on a guessed background.

Workflow

1. Locate the resume. Ask the user to paste it or give a path. If it's a PDF, convert to text first. If they have none, say so and build master-resume.md entirely from the interview in step 3. Never invent a background.
2. Import to documents/master-resume.md. This is the canonical, everything-in record, longer than any tailored resume. Preserve every role, date, and real metric verbatim. Use templates/resume-template.md for structure but do NOT cut content here. The master resume keeps everything so resume-tailor has raw material to select from. Flag anything ambiguous (undated role, vague scope) for the user to confirm.
3. Interview to fill gaps, in one batched round, then execute. Ask only what's missing. Prioritize:
   • Quantified achievements. For each recent role: "What changed because you were there? Put a number on it." Revenue, %, headcount, time saved, scale (users/requests/$). If they don't know a number, mark it [metric TK, user to confirm], never guess.
   • Constraints & non-negotiables. Comp floor (the walk-away number), location/remote, visa/work-authorization, earliest start, industries or company types to avoid, management vs IC.
   • Positioning. "In one sentence, what are you the person for?" and "What are the 2–3 proof points behind that?"
   • Search parameters. Target titles (and the adjacent titles that mean the same job under different names), seniority band, company stage/size preference. Batch these into one message, wait for answers, then write the files. Don't drip questions across many turns.
4. Write profile/target-roles.md (see Output). Include title variants, since the same job is posted as "PMM", "Product Marketing Lead", and "Sr. Product Marketer", so job-sourcing searches wide.
5. Write profile/target-companies.md. Two parts. (a) A company-ICP rubric, the criteria that make ANY company a fit (stage, size, industry, business model, signals like recent funding / new exec, product launch, hiring surge). (b) A tiered named list A/B/C of specific companies. It's fine to start with 5–10 named accounts. The rubric is what lets job-sourcing find more.
6. Write profile/profile.md. The positioning statement, top 5–8 quantified achievements, skills inventory, and hard constraints block. This is the short, dense file most skills read first.
7. Optional public enrichment. With the user's OK, web-search their name + GitHub/portfolio/past companies to surface achievements they undersold. Propose additions. The user confirms before anything enters a file. Every claim must be independently true, with no inference presented as fact.
8. Confirm & handoff. Show the user the constraints block and target list, get a thumbs-up, then point them to job-sourcing.

Output

• Writes to:
  • workspace/documents/master-resume.md – canonical full resume, source of truth
  • workspace/profile/profile.md
  • workspace/profile/target-roles.md
  • workspace/profile/target-companies.md
• Format:
  • profile.md. ## Positioning (1–2 sentences), ## Top Achievements (5–8 bullets, each with a real metric or a [TK] marker), ## Skills (grouped), ## Hard Constraints (comp floor, location, authorization, start date, must/never), ## Links.
  • target-roles.md. ## Titles (primary + variants), ## Seniority, ## Must-haves, ## Nice-to-haves, ## Deal-breakers, ## Comp floor, ## Location / remote, ## Timeline.
  • target-companies.md. ## Company ICP (the rubric, as scannable criteria), ## Tier A / ## Tier B / ## Tier C (named companies with a one-line why).

Rules & quality bar

• No fabrication, ever. Achievements, titles, dates, numbers, and credentials come from the user. Missing metrics get a [metric TK] marker that stays visible. The human fills it, not you.
• master-resume.md is complete. Tailored resumes are edited-down. Never trim master-resume for length. That is resume-tailor's job on a per-role copy.
• Constraints must be decision-ready. A comp floor is a specific number, not "competitive". Location is "Remote (US) or NYC hybrid ≤2 days", not "flexible". Vague constraints break downstream scoring.
• Title variants are mandatory. Under-listing titles is the #1 reason a search misses real roles.
• One batched interview. Gather everything, then write. Don't produce a half-built profile on guessed context and don't interrogate across ten turns.
• The user approves before handoff. Setup ends with the user confirming the profile reflects them.

Related skills

• Feeds: everything. job-sourcing, job-fit-scoring, resume-tailor, cover-letter, company-research, networking-and-referrals, story-bank, interview-prep all read profile/ and/or documents/master-resume.md.
• Hands off to: job-sourcing (find roles against the targets you just defined). Consider story-bank next to turn the achievements here into interview-ready stories.
• Routed here by: search-and-scale-os (the orchestrator) whenever profile/ is empty.