Skills from inference-sh-2/skills

This skill runs user-supplied Python in an online sandbox (`infsh/python-executor`) and returns files saved to `outputs/`. It instructs running a piped install `curl -fsSL https://cli.inference.sh | sh` and includes examples calling `https://example.com` and `https://api.github.com`.

High-risk skill for web search and extraction. Installs and runs `https://cli.inference.sh` via `curl -fsSL https://cli.inference.sh | sh` and directs `infsh` network operations plus `infsh login`, which enables remote execution and credential use.

High-risk skill: instructs downloading and executing a remote installer via `curl -fsSL https://cli.inference.sh | sh` and running `infsh login`. It then directs networked CLI commands to `inference.sh` (e.g., `infsh app run ...`) for AI video generation.

High-risk skill: Python SDK documentation includes unsafe examples that execute untrusted code via `eval(call.args["expression"])` and instruct reading `INFERENCE_API_KEY`. It also demonstrates outbound calls to `https://hooks.slack.com/...` and file access like `/path/to/image.png`.

Dangerous skill that downloads and executes a remote installer and runs a CLI. It instructs `curl -fsSL https://cli.inference.sh | sh`, `infsh login`, and `infsh app run`, which perform network calls to `https://inference.sh`.

Documents the `inference.sh` CLI for running AI apps and shows installing via `curl -fsSL https://cli.inference.sh | sh` and using `infsh` commands. It explicitly instructs executing a remote installer (`curl ... | sh`) and storing credentials in `INFSH_API_KEY`, enabling network installation and secret usage.

Dangerous skill: instructs executing a downloaded shell script (`curl -fsSL https://cli.inference.sh | sh`) and running the `infsh` CLI to perform networked Twitter/X actions. It also directs `infsh login` and multiple `https://inference.sh` endpoints, enabling remote execution and credential use.

High-risk skill: JavaScript/TypeScript SDK documentation for `inference.sh` showing client usage, agents, file uploads, proxies, and webhook/tool examples. It enables arbitrary code execution via `eval` and `internalTools().codeExecution(true)`, reads `process.env` secrets like `INFERENCE_API_KEY`, and calls external endpoints such as `https://hooks.slack.com/...`.