community-ops-automation-workflow

Community Ops Automation Workflow

Overview

Use this skill to turn repeated community-ops work into a safe automation workflow. The default model is:

repeated task -> durable input -> automated collection/draft/check
-> human approval gate -> bounded execution -> read-back verification
-> state writeback -> review/update

Automation should reduce repeated labor and improve review quality. Keep public, financial, permission, reward, role, and official-claim actions under explicit human approval.

Read platform-agent-patterns.md when the workflow spans an operator, AI skill, scheduler, bot, API, or platform integration.

Project Inputs

Get the exact repeated action, input source, durable source of truth, tools or platform adapters, existing permission scope, trigger or frequency, approval owner, expected output, completion evidence, retry limit, and stop conditions from the user or approved project documentation.

The skill supplies the handoff and verification method. It does not include a configured bot, scheduler, API, channels, permissions, credentials, or standing approval. Never borrow integrations or targets from examples. Do not request or store tokens, webhook URLs, passwords, or unnecessary private data; refer to existing secure configuration by name or controlled path.

Workflow

1. Name The Repeated Task

Start by identifying the actual operator action:

• collect community signals;
• prepare a daily or weekly report;
• draft activity/reminder/result copy;
• compare skill versions or stale modules;
• update a local status board;
• prepare a creator or high-value user follow-up list;
• check risk, permissions, rewards, or benefit claims.

Do not automate a vague goal such as "increase engagement." Turn it into a concrete repeated task first.

2. Split The Workflow

For each task, define:

• inputs: channels, posts, files, reports, metrics, user-provided notes, or prior skill files;
• source of truth: the durable record that owns the latest confirmed state;
• allowed automation: read, collect, summarize, classify, draft, compare, prepare reminders, create non-destructive local archives, or produce a checklist;
• human gate: what the owner must review before execution;
• execution roles: what the operator, AI/skill, scheduler, and platform adapter each do;
• current state: observed, prepared, pending approval, approved, executed, verified, recorded, blocked, ambiguous, cancelled, or data gap;
• output: report, draft, checklist, version diff, action list, or archive note;
• evidence: IDs, links, timestamps, counts, source files, or "data missing";
• success criteria: what must be true for the workflow to be complete;
• stop/question condition: when to stop, ask, or return a structured data gap;
• retry limit: which transient failure may be retried and how many times;
• failure mode: missing source, stale data, ambiguous target, permissions risk, or public-action risk.