A moderation policy fails when it's vague ("be respectful") or applied inconsistently — members can't predict what's allowed, and mods burn out making judgment calls with no backing. This skill writes a policy that is enforceable: concrete rules with examples, a ladder that matches consequence to behavior, and the process that makes enforcement feel fair even to the person on the receiving end.
Working from a brief
Given the community (platform, size, purpose, audience), write the full policy — tune severity and tone to the space (a professional Slack ≠ a gaming Discord). Keep the public-facing rules short and human; put the operational detail in the moderator guidelines.
Required Inputs
Ask for (if not provided, else infer and label the assumption):
- The community — platform, rough size, purpose, and audience norms
- The values / what "good" looks like here, and the behaviors you most want to prevent
- Team — how many moderators, volunteer or staff, tools available
- Legal/brand constraints — platform ToS, regulated topics, company brand line
Output Format
Code of conduct (public-facing)
Short, plain, and specific. The core rules stated positively where possible, each with a one-line example of what crosses the line so it's not ambiguous. Cover the usual: harassment, hate speech, spam/self-promo, off-topic, NSFW, illegal content — scoped to this community.
Enforcement ladder
Consequence matched to behavior and repetition:
| Level | Trigger | Action | Who can apply |
|---|---|---|---|
| 1 | first minor / borderline | friendly warning, edit/remove | any mod |
| 2 | repeat / clear violation | formal warning, temp mute | any mod |
| 3 | serious or repeated | temp ban (e.g. 7 days) | senior mod |
| 4 | severe / incorrigible | permanent ban | admin |
| Zero-tolerance | threats, doxxing, CSAM, targeted harassment | immediate ban + report | admin, no warning |
Appeals process
How a member contests an action: where to appeal, who reviews (not the acting mod), the timeline, and what can/can't be overturned. Appeals are what make the ladder feel legitimate.
Moderator guidelines (internal)
- Consistency — decide by the rule, not the person; document every Level 2+ action.
- Conflicts — don't moderate a thread you're personally in; hand off.
- Edge cases — borderline calls, sarcasm/context, brigading and coordinated behavior, when to lock vs remove.
- Transparency — what's communicated to the member vs. handled quietly; a mod log.
- Mod wellbeing — rotation for heavy content, escalation for threats to mods.
Quality Checks
- Every rule is concrete enough to predict a call, with an example of the line
- The ladder ties specific triggers to specific actions and who may apply them
- Severe cases (threats, doxxing, CSAM) are zero-tolerance and route to reporting
- An appeals path exists and is reviewed by someone other than the acting mod
- Mod guidelines cover consistency, conflicts of interest, and documentation
- Tone and severity fit this community; it aligns with the platform ToS
Anti-Patterns
- Vague rules ("don't be a jerk") with no examples — unenforceable and arbitrary-feeling
- A single "ban" hammer with no graduated steps for minor issues
- Inconsistent enforcement by mood or by who the member is
- No appeals process (every action feels final and unjust)
- Unwritten rules enforced as if everyone knew them
- Warning-laddering genuine threats or doxxing instead of acting immediately