| name | hypotheses |
| version | 1.0 |
| description | Hypothesis-driven job search: define what you're testing, execute outreach and applications, validate results, decide to continue or pivot |
| triggers | ["hypoth","гипотез","campaign","кампания","outreach plan","plan outreach","validate","валидир"] |
Hypotheses & Campaigns
Define job search hypotheses, execute outreach campaigns, validate results. Strategy lives in hypotheses/, execution in applications/campaign_*/.
Core Principle
Job search is hypothesis-driven. Each hypothesis is a bet: "I think companies like X need someone like me for role Y."
The system separates strategy (what we're testing and why) from execution (messages sent, responses received, status tracking).
Routing Rules — Who Puts What Where
This is the canonical reference for where artifacts live in the project.
hypotheses/ — Strategy Only
Contains only strategic thinking, no execution files.
| File | Purpose |
|---|
hypotheses_index.md | Master index: all hypotheses with status |
[slug]/hypothesis.md | What we're testing, why, success criteria, validation results |
What does NOT belong here:
- Outreach plans →
applications/campaign_*/
- Message templates →
applications/campaign_*/
- Sent messages →
applications/campaign_*/messages/
- CV snapshots →
applications/campaign_*/cv_sent.md
applications/ — Execution
All day-to-day work lives here. Two types of folders:
app_[company]_[date]/ — Formal Application
Frozen snapshot of CV + cover letter sent to one specific company.
| File | Purpose |
|---|
cv_sent.md | Exact CV text sent (frozen, never changes) |
cover_letter.md | (optional) Cover letter text (frozen) |
Created when: User formally applies to a company via ATS, email, or referral.
campaign_[name]/ — Batch Outreach
Working folder for cold outreach to multiple companies. Everything you need day-to-day in one place.
| File | Purpose |
|---|
outreach_plan.md | Company lists, tracks (A/B), priorities, status tracking, stats — single source of truth for the campaign |
cold_outreach_templates.md | Message templates by domain cluster |
cv_sent.md | CV snapshot for batch (one CV sent to all) |
messages/[company].md | Sent message drafts per company |
Created when: User starts a batch outreach campaign (many companies, one hypothesis).
Why everything in one folder: During execution, you shouldn't need to jump between folders. Open campaign folder → everything is there.
target_roles/ — Role Artifacts Only
Role-specific analysis and generated documents. No outreach, no hypothesis content.
| File | Purpose |
|---|
role_profile.md | Merged JD requirements, domain context |
skills_mapping.md | Achievement-to-requirement mapping + readiness assessment |
cv.md | Living CV (evolves with each new JD) |
cover_letter.md | (optional) Cover letter draft for a specific company |
jd_[company]_[date].md | Full original JD text |
achievements/ — Evidence (Stories)
Source of truth for all claims. Role-agnostic — no skills analysis here.
companies_i_worked/ — Company Profiles
Built FROM achievements. Provide context for story-gathering and CV generation.
Quick Reference: Where Does X Go?
| Artifact | Location |
|---|
| "I think post-raise startups need me" | hypotheses/[slug]/hypothesis.md |
| Success criteria for a hypothesis | hypotheses/[slug]/hypothesis.md |
| Validation results | hypotheses/[slug]/hypothesis.md |
| Company list for cold outreach | applications/campaign_[name]/outreach_plan.md |
| Message template for EdTech cluster | applications/campaign_[name]/cold_outreach_templates.md |
| Actual message sent to Company X | applications/campaign_[name]/messages/[company].md |
| Status: Company X replied | applications/campaign_[name]/outreach_plan.md (status + stats) |
| CV sent to 20 companies in a batch | applications/campaign_[name]/cv_sent.md |
| CV sent to 1 company via ATS | applications/app_[company]_[date]/cv_sent.md |
| Cover letter for formal application | applications/app_[company]_[date]/cover_letter.md |
| JD analysis | target_roles/role_[type]/role_profile.md |
| Skills mapping | target_roles/role_[type]/skills_mapping.md |
| Living CV (role-level) | target_roles/role_[type]/cv.md |
Directory Structure
hypotheses/
├── README.md ← what this directory is, how to use it
├── hypotheses_index.md ← all hypotheses with status
└── [hypothesis_slug]/
└── hypothesis.md ← strategy, criteria, validation
applications/
├── README.md
└── my_data/
├── applications_index.md ← master tracking + conversion funnel
├── app_[company]_[date]/ ← formal applications (one per company)
│ ├── cv_sent.md ← frozen snapshot
│ └── cover_letter.md ← (optional, frozen)
└── campaign_[name]/ ← batch outreach (one per campaign)
├── outreach_plan.md ← companies, tracks, priorities, status + stats
├── cold_outreach_templates.md ← message templates by cluster
├── cv_sent.md ← CV snapshot for batch
└── messages/ ← sent message drafts
└── [company].md
Hypothesis Lifecycle
1. Define
User describes what they want to test. AI creates:
hypotheses/[slug]/hypothesis.md — strategy, positioning, success criteriahypotheses/hypotheses_index.md — adds entry with status 🔬 In progressapplications/campaign_[name]/ — creates campaign folder with outreach_plan.md
What goes in hypothesis.md:
- What we're testing and why
- Positioning (what role we're pitching)
- Source (where companies come from)
- Success criteria (defined upfront, measurable)
- Related roles (links to
target_roles/)
- Related campaign (link to
applications/campaign_*/)
What does NOT go in hypothesis.md:
- Company lists → outreach_plan.md in campaign folder
- Message drafts → messages/ in campaign folder
- JD analysis → target_roles/
2. Execute → see Campaign Execution below
Campaign Execution
Detailed guide for setting up and running a batch outreach campaign.
When to Create a Campaign
A campaign is created when a hypothesis moves from "defined" to "executing". Every campaign belongs to a hypothesis.
One hypothesis can have one campaign (typical) or eventually none (if hypothesis is paused before execution starts).
Campaign Setup (Step by Step)
Step 1: Create folder structure
applications/my_data/campaign_[name]/
├── outreach_plan.md
├── cold_outreach_templates.md
├── cv_sent.md
└── messages/
Folder name uses slug + month: campaign_cold_outreach_may2026, campaign_scaling_europe_jun2026.
Step 2: Outreach Plan
outreach_plan.md is the strategic document for this campaign.
Structure:
# Outreach Plan: [Campaign Name]
## Hypothesis
Link to hypothesis.md
## Company Source
Where the list came from (URL, LinkedIn post, job board)
## Tracks
### Track A: Formal Applications
Companies with open JDs that match a defined role.
### Track B: Cold Outreach
Companies where we pitch without an open role.
## Companies by Priority
### Priority 1 (send first)
* Company A — [cluster] — [reason]
* Company B — [cluster] — [reason]
### Priority 2
...
### Do Not Send
* Company X — reason (already applied, no fit, etc.)
Step 3: Cluster Companies
Group companies by domain/industry for template targeting.
Clustering rules:
- Group by what the company does (EdTech, FinTech, HealthTech, Enterprise SaaS, DevTools, etc.)
- Minimum 2 companies per cluster — singletons join the nearest cluster or get a generic template
- Clusters inform template tone and hook (EdTech → education language, FinTech → compliance language)
- One company can belong to only one cluster (pick the primary domain)
Example clusters:
- EdTech / Future of Work: Labster, Bloomtech, Preply
- FinTech / Payments: Rapyd, MoonPay, dLocal
- HealthTech: 9amHealth, Enzo Health, Dandelion Health
- Enterprise / Ops: gaiia, Vori, Nova Intelligence
Step 4: Assign Tracks and Priorities
Track A — Formal Applications:
- Company has an open JD that matches a defined role
- Apply through ATS or official channel
- Use the CV from
target_roles/role_[type]/cv.md
- Create
app_[company]_[date]/ for each
Track B — Cold Outreach:
- No open JD, or JD doesn't match existing roles
- Cold message to founder/CEO/hiring manager
- Use the campaign CV snapshot
- Message goes to
messages/[company].md
Priority assignment:
- Priority 1 — strongest fit, best signal → send first
- Priority 2-3 — good fit, less signal
- Priority 4-5 — stretch fit, experimental
- Do Not Send — bad fit, already applied, conflicting info
Step 5: Write Message Templates
cold_outreach_templates.md contains templates per cluster.
Template structure:
### Cluster: [Name]
**Hook angle:** [what connects user to this domain]
---
**Subject:** [Short, specific]
Hi [First Name],
[1-2 sentences: why this company caught my eye — specific to their product/stage/market]
[1-2 sentences: what I bring — tied to the hook angle, with one concrete metric]
[1 sentence: soft CTA — "happy to chat" / "would love to share how I..." / "open to a conversation?"]
[Sign-off]
---
Template rules:
- One metric per template — not a CV dump, a conversation starter
- Specific to the company's domain — generic "I'm great" messages get ignored
- Under 120 words — respect the reader's time
- Soft CTA — "happy to chat" not "let's schedule a call"
- No attachment mention — LinkedIn message doesn't need it; if sending by email, attach CV
- Personalized per company — template is a starting point, each message gets a specific hook
Anti-patterns:
- ❌ "I'm writing to express my interest in..." — corporate speak
- ❌ "As a [long title] with [X] years of experience..." — CV on first date
- ❌ "I believe my skills would be a great asset to your team" — filler
- ❌ Same message to 20 companies with only name changed — lazy
Step 6: Snapshot CV
Copy the relevant CV from target_roles/role_[type]/cv.md into campaign/cv_sent.md.
If the campaign uses multiple roles (e.g., one CV for Track A, another for Track B), snapshot both and note which is for which track.
Step 7: Execute — Write and Send Messages
For each company in priority order:
- Personalize the template — find specific detail about the company (product launch, funding, market, recent news)
- Write to
messages/[company].md — include subject line + body
- Update
outreach_plan.md — change status in the relevant table + update Stats
- User sends — then confirms for status update
Status Tracking in outreach_plan.md
The outreach plan is the single source of truth for campaign status. It includes:
Stats section (at the top, after hypothesis):
## Stats
* **Sent:** N (Track B) + N (Track A formal)
* **Replied:** N
* **Reply rate:** X%
* **Interviews:** N
Status codes in tables:
| Status | Meaning |
|---|
| ⬜ Ready | Identified, not sent yet |
| ✅ Sent / Applied | Message or application sent |
| ✅ Replied | Positive response received |
| ❌ Rejected | Rejection received |
| ➡️ → Formal app | Cold outreach led to formal application |
| 🚫 Do not send | Decided not to reach out |
| 💬 Conversation | Ongoing conversation |
Tracking Results in applications_index.md
Campaign activity also appears in applications/my_data/applications_index.md:
- Cold outreach messages count in Pipeline → Sent
- Replies count in Pipeline → Replied
- Channel: "Cold outreach"
- Individual entries for each company in the "All Applications" section
Why two places: outreach_plan.md is for campaign-level tracking (all companies in one view). applications_index.md is for overall job search tracking (all channels, all roles, conversion funnel). When updating status, always update both.
Formal Applications Within a Campaign
When cold outreach leads to a formal application:
- Create
applications/my_data/app_[company]_[date]/ with frozen CV + CL
- Update
outreach_plan.md — change status to ➡️ → Formal app + update Stats
- Update
applications_index.md — add entry with channel "Cold outreach → Formal"
- The campaign folder tracks the cold message; the app folder tracks the formal application
One hypothesis can use multiple roles. Example: cold outreach uses "Founding Product & Ops Lead" CV, but a formal application to the same company might use "Customer Success Manager" CV.
3. Validate
When enough data is collected:
- Compare results vs success criteria
- Identify patterns (which clusters responded? which didn't?)
- Document lessons in hypothesis.md (Validation section)
- Update hypotheses_index.md with status change
4. Decide
Based on validation:
- Continue — hypothesis is working, keep going
- Pivot — partial signal, adjust positioning or company selection
- Stop — hypothesis invalidated, archive it
hypotheses_index.md Format
# Hypotheses Index
***
## Status Overview
| Status | Count | Hypotheses |
| -------------- | ----- | ---------- |
| 🔬 In progress | N | Name |
| 📋 Queued | N | — |
| ✅ Validated | N | — |
| ❌ Invalidated | N | — |
| ⏸ Paused | N | — |
***
## Hypotheses
### 1. [Hypothesis Name]
* **Status:** 🔬 In progress
* **Hypothesis:** One-sentence summary
* **Positioning:** Role title
* **Source:** Where companies came from
* **Started:** YYYY-MM-DD
* **Completed:** —
* **Folder:** `[slug]/`
* **Campaign:** `applications/my_data/campaign_[name]/`
hypothesis.md Format
# [Hypothesis Name]
## What We're Testing
[Strategy description]
## Execution Plan
### Track A: Formal Applications
### Track B: Cold Outreach
* **Templates:** `applications/my_data/campaign_[name]/cold_outreach_templates.md`
* **Messages sent:** `applications/my_data/campaign_[name]/messages/`
### Company Clusters
[Groupings by domain]
**Detailed plan + status tracking:** `applications/my_data/campaign_[name]/outreach_plan.md`
***
## Success Criteria
| Metric | Target | Actual |
|--------|--------|--------|
| Response rate | X% | — |
| Interviews | Y | — |
***
## Related Roles
- `target_roles/my_data/role_[type]/` — [role description]
***
## Validation
[Fill when hypothesis concludes]
Integration With Other Skills
- job-analysis: Creates role folders → hypotheses reference them
- map-skills / generate-cv: Produces CVs → campaigns snapshot them
- applications: Formal apps (
app_*/) tracked alongside campaigns (campaign_*/) in the same index
- cover-letters: CLs for formal applications live in
app_*/, not in campaigns
- starr-achievements: Source of truth — CV content comes from here
Data flow:
achievements → target_roles (skills_mapping, cv)
→ companies_i_worked (profiles)
hypotheses (strategy) → applications/campaign (execution)
→ applications/app (formal snapshots)
📎 External References
External materials for this skill: /references/hypotheses/sources.md
