Openai

OpenAI API integration — chat completions, embeddings, image generation, audio transcription, file management, fine-tuning, and assistants via the OpenAI RES...

This skill claims to be a full OpenAI CLI and only asks for your OPENAI_API_KEY, which is reasonable in principle — but the implementation has inconsistencies that raise caution. Before installing or running it: 1) Inspect the script yourself (or ask someone you trust) — it looks like a lightweight wrapper but uses nonstandard API paths and methods and will likely not perform some operations (file uploads, transcriptions) as advertised. 2) Note it will try to read WORKSPACE or ~/.openclaw/workspace/.env to find OPENAI_API_KEY if the environment variable is absent; make sure that file doesn't contain other secrets you don't want read. 3) If you test it, use a limited-scope or expendable API key and small test inputs. 4) Verify the skill's provenance (homepage and author links) and prefer an implementation that uses the official OpenAI endpoints/SDK or clearly documents a proxy. 5) If you need the declared features (multipart uploads, audio file uploads, generation flows), demand a corrected implementation that actually sends files and proper request bodies. If you are not comfortable auditing the code, treat this as untrusted and avoid installing it with a production or high-privilege API key.

Type: OpenClaw Skill Name: openai Version: 1.0.0 The skill bundle is a non-functional facade that claims to provide comprehensive OpenAI API integration but contains logic that is entirely disconnected from the actual OpenAI API. Specifically, `scripts/openai.py` uses incorrect HTTP methods (GET instead of POST for chat/embeddings), targets non-existent endpoints (e.g., `/v1/chat`), and completely ignores user-provided arguments like chat prompts or file paths in its command functions. Furthermore, `SKILL.md` claims support for non-existent models like 'GPT-5'. While no direct evidence of data exfiltration to third-party domains was found (it hardcodes `api.openai.com`), the deceptive nature of the implementation combined with its access to sensitive `OPENAI_API_KEY` credentials from environment variables and local `.env` files makes it highly suspicious.