Community Moderation

Implement fair, consistent moderation that keeps communities healthy and inclusive.

Skill Contract

Parameters

parameters:
  required:
    - action_type: enum[review, warn, remove, ban, appeal]
    - incident_context: string
  optional:
    - severity: enum[minor, moderate, serious, severe]
    - prior_violations: integer

Output

output:
  moderation_action:
    decision: enum[no_action, warning, removal, ban]
    rationale: string
    communication: string

Code of Conduct

Core Principles

1. Be respectful and inclusive
2. No harassment, discrimination, or hate speech
3. Keep discussions constructive
4. Respect privacy and confidentiality
5. Follow platform rules

Enforcement Ladder

Level	Violation	Action	Record
1	Minor (off-topic)	Friendly reminder	30 days
2	Moderate (rude)	Warning + delete	90 days
3	Serious (harassment)	Temp ban (1-7d)	1 year
4	Severe (threats)	Permanent ban	Forever

Moderation Workflow

Report → Review → Decide → Act → Document → Follow-up
   ↓        ↓        ↓       ↓        ↓          ↓
 Flag    Context  Policy   Remove   Log      Appeal
 System  Check    Apply    Warn     Event    Process

Conflict Resolution

De-escalation Steps

1. Acknowledge: "I understand you're frustrated"
2. Separate: Move to private channel if needed
3. Listen: Let both parties share
4. Find Common Ground: Focus on shared goals
5. Resolve: Propose fair solution

Difficult Situations

Situation	Approach	Escalation
Heated debate	Cool off period	Community manager
Repeated offender	Clear warning	Ban review
Doxxing/threats	Immediate ban	Legal team
Gray area	Team discussion	Policy update

Moderation Team Roles

• Community Manager: Strategy, escalations
• Senior Moderators: Complex cases
• Moderators: Day-to-day enforcement
• Champions: Limited powers

Retry Logic

retry_patterns:
  unclear_violation:
    strategy: "Consult with another moderator"
    escalation: "Team vote on action"

  member_dispute:
    strategy: "Private mediation session"
    fallback: "Both parties warned"

  appeal_received:
    strategy: "Different moderator reviews"

Failure Modes & Recovery

Failure Mode	Detection	Recovery
Inconsistent enforcement	Complaints	Audit, recalibrate
False positive ban	Appeal success	Apologize, restore
Mod burnout	Slow responses	Rotate, recruit

Debug Checklist

□ Violation matches CoC section?
□ Context fully understood?
□ Prior history reviewed?
□ Consistent with past cases?
□ Action proportionate?
□ Communication respectful?
□ Documentation complete?
□ Appeal path explained?

Test Template

test_moderation:
  unit_tests:
    - test_violation_classification:
        assert: "Correct severity level"
    - test_response_time:
        assert: "Reviewed within 4 hours"

  integration_tests:
    - test_appeal_process:
        assert: "Different reviewer, 72h decision"

Observability

metrics:
  - reports_per_week: integer
  - resolution_time_avg: duration
  - appeal_rate: float
  - overturn_rate: float

See assets/ for code of conduct templates.