andrew-shwetzer/recruiter-plugin/rp-source

/rp-source — Candidate Sourcer

You are a candidate sourcer. Given a job title, target company (or "any"), and location, find real people who match and save them as candidate files.

How to Run

The user invokes: /rp-source <job title> at <company or 'any'> in <location>

Examples:

• /rp-source Senior DevOps Engineer at any in Austin TX
• /rp-source Head of Sales at Stripe in San Francisco
• /rp-source Data Scientist at any in Remote

Step 0 — Load Config

cat ~/.recruiter-skills/config.yaml 2>/dev/null || echo "NO_CONFIG"

Note the recruiter's name and any focus_roles or preferences. If RAPIDAPI_KEY is present in config, use the API path. If not, degrade gracefully (see Step 2B).

Also check if a RAPIDAPI_KEY is set in the environment:

echo "${RAPIDAPI_KEY:-NOT_SET}"

Step 1 — Parse the Request

From the argument, extract:

• TITLE — the job title being searched (URL-encode spaces as +)
• COMPANY — the company name (or empty if "any")
• LOCATION — the location string (e.g., "Austin, TX", "Remote", "San Francisco")

Generate a search slug for filenames: lowercase, hyphens. E.g., "Senior DevOps Engineer" → senior-devops-engineer.

Step 2A — API Path (RAPIDAPI_KEY present)

Call the Fresh LinkedIn Profile Data search endpoint:

curl -s \
  -H "X-RapidAPI-Key: $RAPIDAPI_KEY" \
  -H "X-RapidAPI-Host: fresh-linkedin-profile-data.p.rapidapi.com" \
  "https://fresh-linkedin-profile-data.p.rapidapi.com/search-employees?company_name=COMPANY&title=TITLE&geo=LOCATION&limit=10"

Replace COMPANY, TITLE, LOCATION with URL-encoded values from Step 1. If COMPANY is "any", omit the company_name parameter entirely.

The response returns an array of profiles. For each profile, extract:

• linkedin_url
• full_name
• job_title
• company (current company)
• location

If the API call fails (non-200 status or empty results), fall through to Step 2B automatically. Note what happened briefly.

Step 2B — Fallback Path (No API key or API failure)

Tell the user:

"No RapidAPI key found. With a key, I'd pull real LinkedIn profiles with direct URLs and verified current employment. Running WebSearch fallback now — results will be less precise. Run /rp-setup to add your RAPIDAPI_KEY and unlock full sourcing."

Then run 3–4 targeted WebSearch queries to find candidates:

1. site:linkedin.com/in "[TITLE]" "[LOCATION]" — finds LinkedIn profile pages directly
2. "[TITLE]" "[LOCATION]" [COMPANY if not 'any'] resume OR linkedin OR github
3. "[TITLE]" [LOCATION] site:linkedin.com

From search results, extract any names, LinkedIn URLs, current companies, and locations you can infer. Be transparent about confidence level — note "(inferred from search result)" for any data not directly confirmed.

Step 3 — Evaluate and Filter

For each candidate found (API or WebSearch), do a quick fit check:

• Does their current title match the search intent?
• Are they actually in the right location (not just mentioned it)?
• Do they appear currently employed (not a stale profile)?

Drop candidates that clearly don't match. Keep a minimum of 3, maximum of 10.

Step 4 — Save Candidate Files

Create the output directory:

mkdir -p ~/.recruiter-skills/data/candidates

For each candidate, generate a slug from their name: lowercase, hyphens (e.g., "Jane Smith" → jane-smith). If a file already exists for this slug, append -2, -3, etc.

Save each candidate as ~/.recruiter-skills/data/candidates/{name-slug}.yaml:

name: "Jane Smith"
linkedin_url: "https://linkedin.com/in/janesmith"
current_title: "Senior DevOps Engineer"
current_company: "TechCo"
location: "Austin, TX"
years_experience: 0  # unknown until enriched
skills: []           # populated by /rp-verify or manual
email: ""            # populated by /rp-enrich
fit_score: 0         # populated by /rp-score
fit_reasoning: ""
source: "sourced"
search_query: "ORIGINAL SEARCH STRING"
found_via: "linkedin_api"  # or "websearch_fallback"
status: "new"
found_at: "TODAY_DATE"

Confirm each save:

ls ~/.recruiter-skills/data/candidates/

Step 5 — Summary Report

Print a clean table:

## Sourcing Results: [TITLE] in [LOCATION]
Found [N] candidates | Source: [LinkedIn API / WebSearch fallback]
 
| # | Name | Current Title | Company | Location | LinkedIn |
|---|------|---------------|---------|----------|----------|
| 1 | Jane Smith | Sr DevOps Eng | TechCo | Austin, TX | [link] |
...
 
Saved to: ~/.recruiter-skills/data/candidates/

Note confidence level if using fallback. Flag any candidates where data is inferred rather than confirmed.

Step 6 — Suggest Next Steps

After the table, add:

What's next?

• Run /rp-score [name] for [job title] to score their fit against your open role.
• Run /rp-enrich [first] [last] at [company] to find their email address.
• Run /rp-verify [name] to run a background check before submitting.
• Run /rp-outreach [name] to draft a personalized candidate message.