Apply Pipeline

After Stage 1, show this dashboard:

=== STAGE 1 COMPLETE: EVALUATION ===

  Company:        [name]
  Role:           [title]
  Location:       [location]
  H1B:            [status]
  Match:          ~[X]%
  Recommendation: [GO / STRETCH / PASS]
  Priority:       P[X] — [category]

  Strengths: [2-3 key strengths]
  Gaps:      [1-2 gaps]

  → Proceed to Stage 2 (Resume Tailoring)?
  → Or: ask me questions, give feedback, or say "pass" to stop here.

If PASS: Explain why briefly and stop. Don't push applying to something that's not a fit.

If GO or STRETCH: Wait for your confirmation before proceeding.

STAGE 2: TAILOR

Execute the full /tailor skill logic:

1. Read jamie/content_library.md and jamie/resume.md
2. Select best variant emphasis for each role
3. Pick 4 bullets per role
4. Fine-tune wording (not cliche!)
5. Check one-page constraint

After Stage 2, show this dashboard:

=== STAGE 2 COMPLETE: RESUME TAILORING ===

  Emphasis:       [variant set chosen]
  Changes made:   [X] of 20 bullets modified
  Self-intro:     [which version recommended]

  CHANGES SUMMARY:
  • Transition Projects: [unchanged / bullet X swapped / reworded]
  • InGenius Prep:       [which emphasis set, what changed]
  • NextGen Healthcare:  [unchanged / minor tweaks]
  • Vestas:              [unchanged / minor tweaks]
  • Kronos:              [unchanged / minor tweaks]

  ACTION ITEMS FOR JAMIE:
  □ Open resume in [Google Docs / Word]
  □ Make the [X] bullet changes listed above
  □ Check it still fits on one page
  □ Export new PDF

  → Ready for Stage 3 (Outreach)?
  → Or: "this sounds too cliche", "change bullet X", "show me the full bullets"

FEEDBACK LOOP: This is where you will most likely push back. Common responses:

• "Too cliche" → revert to closer-to-original wording, just reorder bullets
• "I like bullet X better from the other set" → swap it in
• "That's too long" → trim the specific bullet
• "Show me side by side" → show current vs. proposed for each role

Stay in Stage 2 until you say you're satisfied. Don't rush to Stage 3.

STAGE 3: OUTREACH

Execute the full /outreach skill logic:

1. Search for contacts (alumni, team members, hiring managers)
2. Check existing connections in tracker
3. Draft personalized messages
4. Build outreach sequence

After Stage 3, show this dashboard:

=== STAGE 3 COMPLETE: OUTREACH ===

  Contacts found: [X]
  Alumni:         [names or "none found"]
  Team members:   [names or "none found"]
  Hiring manager: [name or "not identified"]

  OUTREACH SEQUENCE:
  1. [Today]     → Send LinkedIn request to [Contact 1] (reason)
  2. [Today]     → Send LinkedIn request to [Contact 2] (reason)
  3. [Tomorrow]  → Submit application
  4. [Day 2-3]   → Follow up with accepted connections
  5. [Day 3-5]   → Email hiring manager with resume attached

  ACTION ITEMS FOR JAMIE:
  □ Review messages below — tell me if tone is right
  □ Send connection request to [Contact 1]
  □ Send connection request to [Contact 2]
  □ [If email] Send email to [Contact 3]

  → Ready for Stage 4 (Application Prep)?
  → Or: "rewrite message for X", "too eager", "find more people"

FEEDBACK LOOP: Messages are personal — you will want to adjust tone. Listen for:

• "Too formal" / "too casual" → adjust register
• "Don't mention X" → remove that reference
• "Can you find someone in [specific team]?" → do another targeted search
• "This sounds like I'm begging" → make it more peer-to-peer, less requesting

STAGE 4: APPLICATION PREP

This is the pre-flight checklist before Jamie hits "Apply."

=== STAGE 4: APPLICATION PREP ===

  PRE-FLIGHT CHECKLIST:
  ✅ Job evaluated — [GO / STRETCH] with [X]% match
  ✅ Resume tailored — [X] bullets customized
  ✅ Outreach — [X] contacts identified, messages drafted
  □  Export your tailored resume as PDF
  □  Cover letter needed? [Yes — draft below / No / Optional]

  COVER LETTER (if needed):
  [Draft using content_library.md building blocks — self-intro + why this company
   + 2-3 pillars connecting your experience to JD + closing]

  APPLICATION ANSWERS (common questions):
  • Willing to relocate? → [Yes — moving to Portland / Already in Portland / Open to relocation]
  • Salary expectations? → $80K-$130K (adjust based on role level and location)
  • Visa sponsorship needed? → Yes, H1B sponsorship required
  • Years of experience? → ~3 years professional experience
  • Highest education? → MS Applied Organizational Psychology, USC (2023)

  FINAL ACTION ITEMS:
  □ Export tailored resume as PDF
  □ Send outreach messages (LinkedIn)
  □ Submit application on [company ATS]
  □ Log in Google Sheets tracker: date, company, role, status = "Applied"
  □ Set reminder: follow up in 1 week if no response

  → All done! Good luck! 🙏

ACTIVE LISTENING & FEEDBACK PROTOCOL

Throughout the ENTIRE pipeline, watch for your feedback and adapt immediately:

Tone Feedback

Process Feedback

Learning Feedback (save for future)

If you give feedback that should apply to ALL future applications (not just this one), acknowledge it clearly: "Got it — I'll remember [X] for future applications too." This includes preferences about tone, bullet style, types of roles you don't want, etc.

MODEL SELECTION — Use the Cheapest Model That Can Do the Job (v1.1)

This pipeline spawns background agents at Stages 1, 3, and sometimes 4. Always specify model: explicitly — unspecified defaults to Sonnet and wastes budget.

Rule: Any agent doing pure data retrieval or writes → haiku. Any agent drafting words or making fit judgments → sonnet.

TOKEN EFFICIENCY — Baked-In Rules (v1.0)

David and Jamie share a Claude Team plan with per-seat limits (~55K tokens per 5-hour window). Every skill must be as token-efficient as possible without sacrificing quality.

File Read Strategy: Compact First, Expand on Demand

Session Hygiene (automatic)

• After completing a full pipeline run, suggest /clear to Jamie before next job
• Never carry forward stale context from a previous evaluation into a new one
• If Jamie evaluates multiple jobs in sequence, each one should be independent

Enrichment: On-Demand Only

• Do NOT auto-generate cover letters or contact research unless Jamie asks or says "go ahead"
• Stage 3 (outreach) and Stage 4 (apply) should ask before doing expensive work:
  • "Want me to search for contacts at [Company]?" (Chrome browsing = expensive)
  • "Should I draft a cover letter?" (Sonnet generation = expensive)
• If you say "just prep the checklist" → skip contact search + cover letter, go straight to checklist

Sub-Agent Rules

• Every spawned agent MUST specify model: "haiku" unless it needs voice/judgment
• Chrome navigation agents: sonnet (needs page understanding)
• Notion CRUD / file lookup agents: haiku
• If in doubt, use haiku — it handles 90% of mechanical tasks correctly

OUTPUT PRINCIPLES

1. Dashboards at every stage — you should always know where you are and what's next
2. Clear action items — every dashboard ends with specific checkboxes for you
3. Progress indicators — update the pipeline overview as stages complete
4. Summaries, not walls of text — lead with the verdict, details below
5. Ask before proceeding — never auto-advance to the next stage without confirmation
6. Be honest about gaps — you trust directness over encouragement