tw fashion social manager

台灣服飾品牌的社群媒體與 Meta 廣告策略專家。基於 47,000+ 則真實廣告數據，協助規劃貼文內容、廣告文案、標籤組合、發佈時間、三層漏斗策略與台灣在地化行銷。當使用者提到任何下列需求時，都應主動使用此 skill：寫貼文、寫廣告文案、規劃社群內容、選標籤、排發佈時間、設計廣告漏斗、Meta/IG/FB/T...

This skill appears to do what it says (fetch Meta/Threads insights and produce social strategy), but there are important inconsistencies and privacy risks you should consider before installing: - The package metadata declares no required env vars, yet the README and script require META_ACCESS_TOKEN, META_PAGE_ID, META_IG_ACCOUNT_ID and THREADS tokens/credentials. Ask the author to declare these env vars explicitly in the skill metadata so you know what secrets are needed. - The SKILL.md uses hard-coded, user-specific paths (e.g., /Users/shangweilin/projects/openclaw/.env and a PYTHONPATH in your home). Before running, update those commands to point to a safe, project-specific .env or pass tokens explicitly; do not let the skill blindly read an existing .env in your repo root. - The README suggests creating a ‘permanent’ system user token. Prefer least-privilege and short-lived tokens where possible; avoid storing long-lived master tokens in a shared .env. - The skill will write reports and update agent-managed files under ~/.openclaw — review and sandbox this behavior if you are concerned about data leakage or accidental overwrites. - If you plan to run the included script, inspect the code (scripts/fetch_insights.py) yourself (it is readable) and run it in an isolated environment. Confirm the API endpoints are only graph.facebook.com and graph.threads.net and that no unexpected external hosts are contacted. Recommended actions: request the author correct metadata (explicit required env vars), replace hard-coded paths with variables or instructions to use a safe known path, avoid using permanent tokens if possible, and run the script in a controlled/test account first.

Type: OpenClaw Skill Name: tw-fashion-social-manager Version: 1.0.8 The skill bundle contains instructions in SKILL.md that direct the AI agent to execute shell commands, including package installation and running a script that reads sensitive credentials from a .env file. It features hardcoded absolute file paths (e.g., /Users/shangweilin/...) which indicates poor security hygiene and potential execution failures or path traversal risks. While the Python script fetch_insights.py appears to legitimately interact with Meta and Threads APIs, the pattern of instructing an agent to perform shell operations on sensitive configuration files is a high-risk capability that could be exploited if the agent is prompted to deviate from the script's logic.