Sociclaw

An autonomous social media manager agent that researches, plans, and posts content.

This skill appears to do what it says: planning, content generation, optional image generation, and optional Trello/Notion sync. Before installing, consider: (1) Text‑only features require no env vars, but enabling images/credits or research requires configuring an image API base URL, an API key or a provisioning gateway URL and (optionally) Trello/Notion keys — only set those on hosts you trust. (2) The skill writes local files under .sociclaw and .tmp; if you run it on a shared or production machine, review those files and file locations. (3) The provisioning gateway URL (SOCICLAW_PROVISION_URL) will receive your provider and provider_user_id to create image API keys — only use a gateway you trust. (4) Remote logo/image fetching is disabled by default and local image inputs are restricted to allowlisted directories; these protections reduce accidental data exposure but can be changed by env flags. (5) If you will enable managed topups/provisioning, review the gateway/topup endpoints and policies outside the skill (they are external services). If you want extra assurance, run the skill in an isolated environment (virtualenv or dedicated VM) and inspect the code paths you expect to use (image provisioning, topup, Trello/Notion sync) before enabling those integrations.

Type: OpenClaw Skill Name: sociclaw Version: 0.1.9 The OpenClaw SociClaw skill is classified as benign. The code and documentation demonstrate a strong focus on security and transparency. Key indicators include explicit hardening measures such as restricted local file access for image inputs (`SOCICLAW_ALLOWED_IMAGE_INPUT_DIRS`), strict allowlisting and IP address checks for remote image URLs (`SOCICLAW_ALLOWED_IMAGE_URL_HOSTS`, `_is_allowed_remote_image_url` in `image_provider_client.py`), and a deliberate decision to prevent automatic `git pull` or `pip install` during self-update (`cmd_self_update` in `cli.py`). The `SKILL.md` provides clear agent instructions and guardrails against prompt injection, explicitly detailing local file writes and network calls, all of which align with the stated purpose of social media management. No evidence of intentional data exfiltration to unauthorized endpoints, malicious execution, persistence mechanisms, or obfuscation was found.