Skill: Profile Health Scoring

Purpose

Compute LinkedIn profile health score (0-100) and health state (GREEN/YELLOW/ORANGE/RED) based on account metrics. This is the core decision engine for activity level determination.

What it does

•Accepts profile metrics (account age, connection count, daily app count, restriction history, activity diversity, session patterns, acceptance rate)
•Applies deterministic weighted scoring algorithm
•Returns health score, state classification, factor breakdown, and recommended actions
•Logs every calculation for audit trail and analysis

Why it matters

Profile health determines operational constraints:

•GREEN (85-100): Full operations, no restrictions
•YELLOW (60-84): Reduced activity, max 2 apps/day
•ORANGE (30-59): Hydration only, no applications
•RED (0-29): Quarantine, zero activity

How to use it

Single profile health check

python

from scoring.health_calculator import calculate_health

result = calculate_health({
    "profile_id": "P-001",
    "account_age_days": 180,
    "connection_count": 350,
    "daily_application_count": 2,
    "days_since_last_restriction": 45,
    "activity_diversity_score": 85,
    "session_pattern_regularity_score": 90,
    "connection_request_acceptance_rate": 75
})

print(f"Profile {result['profile_id']} health: {result['score']} ({result['state']})")
print(f"Recommended actions: {result['recommended_actions']}")

Portfolio health snapshot

python

from scoring.health_calculator import calculate_portfolio_health

portfolio = calculate_portfolio_health(profile_list)

print(f"Portfolio average health: {portfolio['average_health_score']}")
print(f"State distribution: {portfolio['state_distribution']}")
print(f"Critical actions needed: {portfolio['immediate_actions']}")

CLI usage

bash

# Demo with sample data
python health_calculator.py

# Single profile from JSON
python health_calculator.py profile_data.json

# Batch portfolio from JSON array
python health_calculator.py portfolio_list.json

Input schema

json

{
  "profile_id": "P-001",
  "account_age_days": 180,
  "connection_count": 350,
  "daily_application_count": 2,
  "days_since_last_restriction": 45,
  "activity_diversity_score": 85,
  "session_pattern_regularity_score": 90,
  "connection_request_acceptance_rate": 75
}

Field descriptions

•profile_id: Unique identifier for the LinkedIn profile
•account_age_days: How many days since profile was created
•connection_count: Total number of 1st-degree connections
•daily_application_count: How many LinkedIn applications submitted today
•days_since_last_restriction: Days since last LinkedIn warning/restriction (-1 if never restricted)
•activity_diversity_score: 0-100, measure of activity variety (applications, connections, reactions, views, etc.)
•session_pattern_regularity_score: 0-100, how human-like are login/activity patterns (higher = more random, more human)
•connection_request_acceptance_rate: 0-100, % of sent connection requests that are accepted

Output schema

json

{
  "profile_id": "P-001",
  "score": 78,
  "state": "YELLOW",
  "state_definition": {...},
  "breakdown": {
    "account_age": { "raw_score": 100, "weight": 0.15, "contribution": 15 },
    "connection_count": { "raw_score": 100, "weight": 0.15, "contribution": 15 },
    ...
  },
  "recommended_actions": ["reduce_daily_applications_to_2"],
  "weights_version": "1.0.0",
  "next_check": "2026-02-16T12:00:00Z",
  "timestamp": "2026-02-15T12:00:00Z"
}

Weight configuration

Weights are defined in health_weights.json and are human-editable. Current weights:

•Account age: 15%
•Connection count: 15%
•Daily application count: 20% (highest sensitivity to over-activity)
•Days since last restriction: 15%
•Activity diversity: 15%
•Session pattern regularity: 10%
•Connection request acceptance rate: 10%

To adjust weights, edit health_weights.json. All future calculations will use the new weights.

Health state decision logic

GREEN (85-100)

•Healthy account, no risk signals
•Activity allowed: Full tier operations (Tier A/B/C standard activity)
•Applications: Up to tier daily limit (A=5, B=2, C=1)
•Action: Continue normal operations
•Monitor: Weekly health checks

YELLOW (60-84)

•Account showing stress indicators
•Activity allowed: Light activity, max 2 applications per day
•Applications: Forced to max 2/day regardless of tier
•Action: Reduce activity, increase session gaps
•Monitor: Daily health checks

ORANGE (30-59)

•Account at risk, needs recovery
•Activity allowed: Hydration only (browsing, reactions, light engagement)
•Applications: ZERO applications
•Action: Stop all applications, hydration only
•Monitor: Twice daily health checks

RED (0-29)

•Account compromised or severely restricted
•Activity allowed: ZERO activity
•Applications: ZERO applications
•Action: Immediate quarantine, alert human and Z
•Monitor: Continuous investigation

State transitions

Automatic state transitions occur when:

•Score crosses threshold (e.g., GREEN→YELLOW when score drops below 85)
•Hard restrictions detected (e.g., ANY→RED when Trust & Safety alert)
•Daily application limit exceeded (e.g., GREEN→YELLOW when daily apps hit 4+)

Audit trail

Every health calculation is logged to scoring_log.jsonl with:

•Event type
•Timestamp
•Profile ID
•Score and state
•Full breakdown of all factors
•Weights version used

This allows:

•Replay of any calculation
•Trend analysis over time
•Debugging of state changes
•Compliance reporting

Error handling

•Missing input: Returns error with missing field name
•Invalid profile_id: Still processes, logs with "UNKNOWN"
•Extreme values: Clamped to valid ranges, logged as anomaly
•File I/O errors: Raised with clear message

Performance

•Single profile: <10ms
•Portfolio of 100: <500ms
•Batch logging: Append-only, <5ms per entry

Integration points

•Activity Scheduler: Uses health state to determine activity volume
•Application Executor: Uses health state to determine if applications allowed
•Daily Reporting: Includes portfolio health snapshot
•Risk Assessment: Triggers escalation if many profiles RED