Install
openclaw skills install sociclawAn autonomous social media manager agent that researches, plans, and posts content.
openclaw skills install sociclawSociClaw is an AI agent dedicated to managing social media accounts autonomously. Drafts can be synced to Trello/Notion, and images are optional via a configured SociClaw Image API.
🔵Soci:/sociclaw), answer in 3 parts:
/sociclaw briefing) in setup flow if not present..sociclaw/company_profile.md)./sociclaw setup maps to CLI command setup (alias of setup-wizard)./sociclaw reset maps to CLI command reset./sociclaw update maps to CLI command self-update (manual instructions only, no code executed)./sociclaw (without subcommand) should act as a welcome+help entrypoint, not as an error dump.No environment variables are required for text-only planning and content generation.
Required baseline inputs for the setup wizard:
Optional, only ask if user opts in:
If using provisioning flow:
Local files written by default:
.sociclaw/runtime_config.json (setup answers).sociclaw/company_profile.md (brand brain).sociclaw/planned_posts.json (generated plan).sociclaw/memory.db (persistent memory to reduce repetition).sociclaw/generated_images/ (local backups of generated images).tmp/sociclaw_state.json (local provisioned API key + wallet address).tmp/sociclaw_sessions.db (topup sessions)Network calls (only when features are enabled/used):
SOCICLAW_PROVISION_URL): sends {provider, provider_user_id, create_api_key} and receives an API key.SOCICLAW_IMAGE_API_BASE_URL): sends prompts, model name, and optional logo/image input (as image_url and/or image_data_url) to generate images and to manage credits (topups).XAI_API_KEY is configured and research is enabled.Local/remote image input safety defaults:
.sociclaw and .tmp).SOCICLAW_ALLOWED_IMAGE_INPUT_DIRS can widen local image input roots.SOCICLAW_ALLOWED_IMAGE_INPUT_DIRS are ignored unless SOCICLAW_ALLOW_ABSOLUTE_IMAGE_INPUT_DIRS=true.SOCICLAW_ALLOW_IMAGE_URL_INPUT=false) and requires SOCICLAW_ALLOWED_IMAGE_URL_HOSTS allowlist when enabled.Role:
Working mental model (algorithmic brain):
Scoring priorities (practical heuristics):
Operational imperatives:
Content creation protocol:
Style and formatting:
Role:
Algorithm Priorities:
Writing Protocol:
Style Rules:
Planning Rules:
Brand Brain:
nano-banana image generation, require a logo/input image URL or local path from setup or per request.Analysis Mode:
Quality Guardrails:
/sociclawWelcome message + quick help (recommended). If the user is not configured yet, start onboarding.
/sociclaw setupConfigure niche, posting frequency, content language, brand logo URL (for img2img), brand-document info, and integrations.
/sociclaw briefingCapture brand context (tone, audience, keywords, forbidden terms, language, brand doc path) to improve content quality.
/sociclaw plan [quarter]Generate a starter plan by default (14 days x 1 post/day). Use full quarter mode when requested (90 days x 2 posts/day).
/sociclaw generateGenerate today's posts (text + image prompt + image) and attach results to Trello/Notion.
Each generated post is persisted to local persistent memory (.sociclaw/memory.db) so future planning can avoid repetitive topics.
/sociclaw syncForce a sync to Trello/Notion.
/sociclaw statusShow plan progress and integration status.
/sociclaw payStart credits topup flow (returns deposit address and exact USDC amount).
/sociclaw paid <txHash>Claim topup after transfer confirmation.
/sociclaw updatePrint safe, manual update steps for the host.
This skill build does not execute git pull or pip install automatically (to reduce security risk and scanner flags).
/sociclaw resetFactory reset local runtime state (config, local session DB, local brand profile, local provisioned user state, persistent memory DB). Requires explicit confirmation.
SociClaw supports img2img workflows (example: nano-banana). Those models require an input image (logo) to work.
The setup wizard collects brand_logo_url which can be a URL or a local path (restricted by allowlists).
To auto-create users + API keys without exposing your admin secret, deploy a small gateway on your backend (Vercel) and set:
SOCICLAW_PROVISION_URL=https://api.sociclaw.com/api/sociclaw/provision
SOCICLAW_IMAGE_API_BASE_URL=https://<your-image-api-domain>
The gateway keeps the upstream admin secret server-side. End-users never see it.
Optional gateway auth (only if your gateway requires it):
SOCICLAW_INTERNAL_TOKEN=your_internal_token
Optional hardening knobs:
SOCICLAW_ALLOW_IMAGE_URL_INPUT (default: false) controls remote logo URL fallback.SOCICLAW_ALLOWED_IMAGE_URL_HOSTS (required if enabling remote URL input): comma-separated allowlist for remote logo fetch fallback.SOCICLAW_ALLOWED_IMAGE_INPUT_DIRS (recommended): .sociclaw,.tmp paths allowed for local image input.SOCICLAW_ALLOW_ABSOLUTE_IMAGE_INPUT_DIRS (default: false) allows absolute dir entries in SOCICLAW_ALLOWED_IMAGE_INPUT_DIRS.If you don't want provisioning, you can run images with a single API key:
SOCICLAW_IMAGE_API_BASE_URL=https://<your-image-api-domain>
SOCICLAW_IMAGE_API_KEY=your_sociclaw_image_api_key
SOCICLAW_IMAGE_MODEL=nano-banana
You can install skills by cloning this repo into your OpenClaw skills folder.
Typical locations:
~/.openclaw/skills (global)<your-workspace>/skills (workspace-local)Example:
git clone https://github.com/sociclaw/sociclaw.git ~/.openclaw/skills/sociclaw
Install/update:
git -C ~/.openclaw/skills/sociclaw pull --ff-only
Then start OpenClaw and run:
/sociclaw
cd D:\sociclaw
python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r requirements.txt
.\.venv\Scripts\python.exe -m pytest -q