# Community Desk

> Paste what your product is (who it is for, who the ideal member is, who runs the community and for how long) and your notes on one online community - a Discord server, a Slack workspace, a Circle space, a subreddit, a Facebook group or a forum. The browser computes the skill's health metrics and flags the warning signs. A paid audit lane says whether the community is healthy, at risk or unhealthy, where members drop off and the top 3 priorities; a paid plan lane writes the core loop, the welcome post and message, the ritual calendar and the power-user ask.

URL: https://community-desk.skillsafe.ai/
API tutorial: https://community-desk.skillsafe.ai/api.html

## Who it is for

Founders, community managers and marketers running one community for a product or brand, who want to know whether it is working, why members go quiet, and what to change this month - grounded in their own numbers rather than generic advice.

## The free metrics (in the browser, no model)

- The community notes as labelled lines ("Members: 2,400", "Monthly active: 610", "Daily active: 70", "New members last 30 days: 180", "New members who posted within 7 days: 14", "Posts last 30 days: 520", "Posts with at least one reply: 290", "Posts by our team: 310", "Posts by the top 5 members: 200", "Median time to first reply: 31 hours") or prose; percentages are used as stated.
- The member band and the owner's phase role (0-100 Community Architect, 100-1,000 Community Manager, 1,000+ Community Enabler).
- The skill's five health metrics: DAU/MAU ratio, new member post rate, thread reply rate, lurker ratio, content created by non-staff; plus the top-5 share of posts and the time to a first reply.
- Flags CM-01..CM-13: most posts from the team, questions waiting more than 24 hours, 5 people making 80%+ of the posts, DAU/MAU under 20%, new members stopping after their intro, no metric computable, no goal, no ideal member, no new member journey, no ritual, no power-user recognition, numbers that do not add up, under 20 founding members.
- Exports: the metrics as Markdown and CSV.

## The paid lanes (one system prompt, routed on `task`)

- `audit`: healthy, at risk, unhealthy or pre-launch (healthy is not allowed when the page raised a warning flag); the stage and phase role; the community model the goal points to (support-driven, product-development, education/enablement, founder-led, or a blend); the shared identity, quoted; where members drop off, quoted; the four warning signs in order, yes wherever the page flagged one; the top 3 priorities, each with something to do today and a health metric.
- `plan`: the goal; the weekly core loop; a pinned welcome post (at most 120 words); a welcome message opening "Hi [Name]," (at most 80 words); 2 or 3 first-week prompts; 2 to 5 rituals with at least one weekly; a power-user program with 1 to 3 offers and a personal ask (at most 80 words); how to close the loop on member feedback; 3 to 5 health metrics with targets only where the input states them (or the skill's 20% for DAU/MAU). With an audit handed over, the plan answers each of its three priorities.

Every reply is checked against the input: quotes word for word, the health rules, the warning signs, the word limits, the ritual cadence, and no invented count, rate, target or duration.

## Source

Derived from the agent skill @coreyhaines31/community-marketing (https://skillsafe.ai/skill/@coreyhaines31/community-marketing), by Corey Haines, in coreyhaines31/marketingskills (MIT): https://github.com/coreyhaines31/marketingskills/tree/main/skills/community-marketing - its community strategy principles (shared identity, value to members first, the flywheel), the playbooks (launching from zero, growing an existing community, the advocate program, community-led support), the health metrics and warning signs, and the output formats; and its reference references/community-models.md: the five community models and the scaling-phase role shift.

Runs on gpt-terra. Part of SkillSafe (https://skillsafe.ai/).