network-scan

Files (8)

• SKILL.md
• browser-setup.md · references/
• data-directory.md · references/
• fit-scoring.md · references/
• prerequisites.md · references/
• priority-hierarchy.md · references/
• evaluate-company.md · scripts/
• resolve-careers.md · scripts/

SKILL.md (first lines)

# Network Scan Skill

> **Priority hierarchy**: See `references/priority-hierarchy.md` for conflict resolution.

Proactively check whether companies where you know someone are hiring for roles that match you. First run builds a cache of company careers page URLs. Subsequent runs reuse the cache, making weekly checks fast.

## File Structure

```
scripts/
  resolve-careers.md     # Subagent for resolving a batch of company careers URLs
  evaluate-company.md    # Subagent for scanning a batch of companies' open roles
```

User data (stored at ~/.proficiently/):
```
~/.proficiently/
  resume/                # Your resume PDF/DOCX
  preferences.md         # Job matching rules
  profile.md             # Work history from interview
  linkedin-contacts.csv  # LinkedIn contacts export
  company-careers.json   # Cached company careers URLs
  network-scan-history.md # Running log of scan results
  jobs/                  # Per-job application folders
```

---

## Workflow

### Step 0: Check Prerequisites

Resolve the data directory, then check prerequisites per `references/prerequisites.md`. Resume, preferences, and linkedin-contacts.csv are all required.

Load these files for use in later steps:
- `DATA_DIR/preferences.md` (target roles, must-haves, dealbreakers, nice-to-haves)
- `DATA_DIR/resume/*` (candidate profile)
- `DATA_DIR/profile.md` (work history, if it exists)

### Step 1: Select Contacts & Extract Companies

Parse `$ARGUMENTS`:
- If a number (e.g., `50`): use that as the contact limit
- If `all`: use all contacts (warn user this may be slow if > 200)
- If empty/missing: default to 25

Read `~/.proficiently/linkedin-contacts.csv`. Sort by "Connected On" descending (most recent first). Take the first N contacts based on the limit.

Extract unique company names from the selected contacts. Skip companies with empty or blank names.

Group contacts by company into a lookup:
```
{
  "Google": [{"name": "Jane Smith", "position": "PM Director", "url": "https://linkedin.com/in/janesmith"}, ...],
  "Stripe": [{"name": "John Doe", "position": "Eng Manager", "url": "https://linkedin.com/in/johndoe"}]
}
```

Report to user: "Found X unique companies from Y contacts. Checking careers pages..."

### Step 2: Resolve Careers Pages (Parallelized)

Load `~/.proficiently/company-careers.json` if it exists (the cache). If it doesn't exist, start with an empty object.

Split companies into three groups:
- **Cached (fresh)**: `last_checked` within last 7 days - use as-is, no work needed
- **Cached (stale)**: `last_checked` older than 7 days - needs re-verification
- **Uncached**: not in cache, or `type` is `"not_found"` and stale - needs full resolution

Report: "X companies from cache, Y need resolution..."

**Parallel resolution using subagents:**

Take all companies needing resolution (stale + uncached) and split them into batches of 10. Spawn one subagent per batch using the Task tool (`subagent_type: "general-purpose"`). Run all batches in parallel.

Each subagent receives:
- A batch of company names to resolve
- Instructions from `scripts/resolve-careers.md`

Each subagent uses `WebSearch` (NOT the browser) to find careers pages:
1. Search: `"[Company Name]" careers jobs site:[company domain if known]`
2. From the search results, identify the careers/jobs page URL
3. Classify the URL type:
   - `"direct"` - company's own careers page (e.g., careers.google.com)
   - `"greenhouse"` - Greenhouse ATS (boards.greenhouse.io/company or company.greenhouse.io)
   - `"lever"` - Lever ATS (jobs.lever.co/company)
   - `"workday"` - Workday ATS (company.wd5.myworkdayjobs.com)
   - `"other_ats"` - other ATS platforms (Ashby, BambooHR, etc.)
   - `"not_found"` - no careers page could be found (set `careers_url` to null)
4. Return results for the batch

Collect results from all subagents and merge into the cache. Save `~/.proficiently/company-careers.json`. Format:
```json
{
  "Company Name": {
    "careers_url": "https://careers.example.com",
    "type": "direct",
    "last_checked": "YYYY-MM-DD",
    "last_found_roles": 0
  }
}
```

Report progress: "Resolved X new careers pages, Y from cache, Z not found."

### Step 3: Scan for Matching Jobs (Parallelized)

Take all companies with a valid `careers_url` (skip `not_found` and `ignored` entries). Split them into batches of 5 companies each.

**Spawn parallel subagents** using the Task tool (`subagent_type: "general-purpose"`). Run all batches in parallel (up to 5 concurrent subagents to avoid overwhelming the browser).

Each subagent receives:
- A batch of companies (name, careers_url, ATS type, network contacts)
- Candidate profile summary (from resume)
- Preferences (target roles, must-haves, dealbreakers, nice-to-haves)
- Instructions from `scripts/evaluate-company.md`

Each subagent:
1. Creates its own browser tab (`tabs_context_mcp` then `tabs_create_mcp`)

From Infrasity-Labs/dev-gtm-claude-skills, MIT.

More from Infrasity-Labs/dev-gtm-claude-skills