PrivateInvestigator

Customization

Before executing, check for user customizations at: ~/.claude/LIFEOS/USER/CUSTOMIZATIONS/SKILLS/PrivateInvestigator/

If this directory exists, load and apply any PREFERENCES.md, configurations, or resources found there. These override default behavior. If the directory does not exist, proceed with skill defaults.

🚨 MANDATORY: Voice Notification (REQUIRED BEFORE ANY ACTION)

You MUST send this notification BEFORE doing anything else when this skill is invoked.

1. Send voice notification:

   curl -s -X POST http://localhost:31337/notify \
     -H "Content-Type: application/json" \
     -d '{"message": "Running the WORKFLOWNAME workflow in the PrivateInvestigator skill to ACTION"}' \
     > /dev/null 2>&1 &
   

2. Output text notification:

   Running the **WorkflowName** workflow in the **PrivateInvestigator** skill to ACTION...
   

This is not optional. Execute this curl command immediately upon skill invocation.

PrivateInvestigator - Ethical People Finding

What It Does

Finds people and verifies identities using public data only. It covers people-search aggregators, social media, public records, and reverse lookups in parallel, then scores results by confidence (HIGH/MEDIUM/LOW/POSSIBLE) and trusts a match only when 3+ independent identifiers align.

The Problem

Finding a real person from a name, a phone number, or an email is scattered across dozens of sources — people-search sites, county records, court portals, social platforms, reverse-lookup services — and no single one gives you the answer. Common names make it worse: search "John Smith" and you get thousands of people, most of them the wrong one. Run the searches one at a time and you either give up before you've covered enough sources or you act on a single weak match and contact the wrong person. This skill runs the sources in parallel and refuses to call a match real until several independent identifiers line up.

How It Works

Public data only. No hacking, pretexting, or authentication bypass — every technique here is legal and ethical. The work runs as a parallel investigation across the source categories below; findings get cross-checked and confidence-scored before anything is reported. Choose the fan-out breadth the case needs — a common name across many states wants wide parallel coverage, a rare name in one city wants far less.

Workflow Routing

When executing a workflow, output this notification:

Running the **WorkflowName** workflow in the **PrivateInvestigator** skill to ACTION...

When to Activate

Direct People-Finding

• "find [person]", "locate [person]", "search for [person]"
• "reconnect with [person]", "looking for lost contact"
• "find an old friend", "locate a former coworker"

Reverse Lookup

• "reverse phone lookup", "who owns this email"
• "reverse image search", "find person by username"

Investigation

• "background check" (public data only)
• "what can you find about [person]"
• "research [person]"

Research Strategy

Done = broad, independent coverage. Every source category below is worked, and no key identifier (address, employer, DOB, family, social handle) rests on a single source. Run the categories in parallel — dispatch research agents concurrently in one message — and scale the fan-out to the case: a common name across many states justifies many parallel agents and spelling/location variants; a rare name in one city needs only a few. Independence is the point, not headcount: aggregators that resell the same database count as one source, so weight toward genuinely distinct origins (social the subject created, government records, unrelated aggregators).

Core Capabilities

1. People Search Aggregators

2. Social Media Investigation

• Facebook: Google x-ray searches, mutual friends, groups
• LinkedIn: Boolean search, alumni networks
• Instagram/Twitter/TikTok: Username patterns, cross-platform correlation

3. Public Records

• Voter Registration: Most states publicly available
• Property Records: County assessor/recorder sites
• Court Records: PACER (federal), state court portals, CourtListener
• Business Filings: Secretary of State websites
• Professional Licenses: State licensing boards

4. Reverse Lookup

• Phone: CallerID, NumLookup, carrier lookup
• Email: Epieos, Holehe, Hunter.io
• Image: PimEyes, TinEye, Google/Yandex Images
• Username: Sherlock, WhatsMyName, Namechk

5. Google Dorking

site:linkedin.com "John Smith" "Software Engineer"
site:facebook.com "lives in" "Austin" "marketing"
filetype:pdf resume "Jane Doe" "San Francisco"

Investigation Methodology

Anchor on whatever foundation identifiers the request gives — full name (and variations/maiden names), approximate age or DOB, last known location, context (school, workplace, relationship) — then pull from every source category until identifiers corroborate. Each discovered fact (phone, email, username, address, relative) is itself a new lead to run back through the sources, and every candidate gets a timeline-consistency and cross-source check before it earns a confidence score. The order is yours; the finish line is a match that survives the 3+-independent-identifier bar below.

Confidence Scoring

Dealing with Common Names

1. Add Specificity - Include location, age, employer, school
2. Cross-Reference - Match DOB + address patterns across sources
3. Family Connections - Verify through known relatives
4. Timeline Analysis - Does the life history make sense?
5. Multiple Identifiers - Require 3+ matching data points

Legal & Ethical Boundaries

GREEN ZONE (Allowed)

✅ Search public records (property, court, voter, business) ✅ Access publicly posted social media content ✅ Use people search aggregator sites ✅ Perform reverse lookups on public data ✅ Google dorking with public search operators

RED ZONE (Never Cross)

❌ Access data behind login walls without authorization ❌ Bypass authentication or security measures ❌ Use pretexting or impersonation ❌ Access private databases (credit, financial, medical) ❌ Stalk, harass, or intimidate subjects ❌ Access PI-only databases without license

When to STOP

• If the purpose shifts to harassment or stalking
• If the subject has clearly opted out of contact
• If investigation requires illegal methods
• If you suspect the requestor has malicious intent

Examples

Example 1: Finding an Old College Friend

User: "Help me find my college roommate from 2005, John Smith from Austin"
→ Routes to FindPerson.md
→ Fans out parallel research across the source categories
→ Cross-references people search + LinkedIn alumni + property records
→ Verifies identity through timeline analysis
→ Reports findings with HIGH confidence

Example 2: Reverse Phone Lookup

User: "Who called from 512-555-1234?"
→ Routes to ReverseLookup.md
→ Runs phone through CallerID, NumLookup
→ Cross-references with people search aggregators
→ Reports owner name, location, carrier

Example 3: Social Media Investigation

User: "Find Jane Doe's social media, she's a marketing professional in Denver"
→ Routes to SocialMediaSearch.md
→ LinkedIn Boolean search + Google x-ray
→ Username enumeration if handle discovered
→ Reports all accounts with MEDIUM/HIGH confidence

Related Documentation:

• Complete workflow details in Workflows/ directory
• Integration with Research skill for parallel agent orchestration

Gotchas

• Ethical framework is mandatory. Legitimate purposes only — reconnection, due diligence, safety. No stalking or harassment.
• Wide parallel fan-out can hit rate limits on public records APIs. Stagger launches if services throttle.
• Verify findings across multiple sources. Single-source results are unreliable.

Execution Log

After completing any workflow, append a single JSONL entry:

echo '{"ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'","skill":"PrivateInvestigator","workflow":"WORKFLOW_USED","input":"8_WORD_SUMMARY","status":"ok|error","duration_s":SECONDS}' >> ~/.claude/LIFEOS/MEMORY/SKILLS/execution.jsonl

Replace WORKFLOW_USED with the workflow executed, 8_WORD_SUMMARY with a brief input description, and SECONDS with approximate wall-clock time. Log status: "error" if the workflow failed.