csgclaw

Run the complete three-stage CSGClaw structured-output demo when the user explicitly invokes $csgclaw-interactive-output-demo. Exercise resource links, Codex-style options, recommended and Unicode labels, freeform alternatives, secret input, and agent-directed automatic continuation.

Mandatory skill for provisioning any new CSGClaw agent-backed participant or worker. Use immediately when the user asks to create, add, set up, or provision an agent, robot, worker, or user-facing "bot" (including GitLab, frontend, backend, QA, or other specialized workers), when dispatch needs a missing worker, or when asking which hub template fits. Always template list + match + template get + participant create --type agent --bind create --from-template with --env for secrets. Never create a CSGClaw worker with --type agent unless it binds a real Agent. Never run participant create --bind create without --from-template for a new worker. Do NOT use for task dispatch to existing workers.

Default manager skill for multi-worker CSGClaw orchestration. Create teams and tasks, plan and start work through `csgclaw-cli team`, and keep dispatch, claim, dependencies, and status on server task state.

Configure and troubleshoot CSGClaw Feishu/Lark channel credentials for manager or worker agents. Use when the Manager needs to generate a Feishu bot app creation URL or QR code, collect App ID/App Secret through registration, bind Feishu participants through `csgclaw-cli participant bind`, recreate workers, or debug Feishu messages not reaching CSGClaw workers.

Use this skill when you are a worker operating inside a CSGClaw task execution room. Only act on explicit task dispatch messages, claim the dispatched task, report blocked/completed/failed status through the team CLI, and keep all execution aligned with server task state.

Search, inspect, and install ClawHub-compatible skills with csgclaw-cli skill. Use when the user asks to find skills on opencsg or clawhub.ai, list versions, or install into this sandbox workspace/skills directory.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Codex's capabilities with specialized knowledge, workflows, or tool integrations.

Draft Conventional Commit messages from the current git changes in a local repository. Use when Codex should inspect staged or unstaged diffs, infer the primary change, suggest one or more commit messages, and warn when the changes should be split into multiple commits.

Generate user-facing weekly release updates from a GitHub release range. Use when Codex needs to fetch releases such as `v0.3.1~v0.3.2`, read their release notes, and write a concise bilingual weekly summary in Chinese and English for users, operators, or community readers.

Review a GitHub pull request by repo and PR number, inspect the PR diff, draft severity-ranked review findings, and optionally publish selected comments back to the PR after explicit user confirmation. Use when Codex is asked to review a specific PR, prepare actionable review feedback, prioritize only the highest-value comments for publication, or handle a two-phase review flow with a preview before posting PR comments.

Draft GitHub release notes from selected commits in a local git repository. Use when Codex needs to turn specific `git log` entries, commit SHAs, or revision ranges into a release summary with a short "What's Changed" narrative, grouped highlight bullets, and a full changelog.