Using Model Endpoint Skill (Planned / Not Wired)

A design for an endpoint-scoped inference kernel, together with the trusted provider shim and manifest it would need. No part of this directory is wired into an executable provider path in a running OpenAI4S today: ComputeManager discovers only remote-compute-* providers and supports the BYOC and SSH families, and host.endpoints.* registers and probes endpoints without ever creating the scoped inference kernel described here.

The Skill itself is still discoverable — the loader picks up every skills/<name>/SKILL.md, this one included, so an agent can list and load it through progressive disclosure. What it cannot do is run it. Read SKILL.md as a runbook written ahead of its implementation, and the provider files as dormant assets. Their presence is not evidence that compute_provider({'provider': ...}) works. Until discovery, lifecycle and routing are connected and tested, none of this may be presented as an end-to-end capability that exists.

Install

A Skill is a directory of files, so installing one is copying that directory somewhere an agent looks. With Node 18+ and git on PATH (npm resolves a github: spec through it), and nothing cloned:

npx github:PKU-YuanGroup/OpenAI4S install using-model-endpoint --target claude

--target claude writes to ~/.claude/skills, claude-project to ./.claude/skills, openai4s to <data_dir>/user-skills, and --dir <path> to anywhere you name. The resolved absolute path is printed before anything is written there, and --dry-run stops at that plan. A reinstall refuses to overwrite a copy you have edited or one it did not install, and uninstall removes only the files it wrote. For a recipe included in the npm release, npx @pku-yuangroup/openai4s-skills install using-model-endpoint --target claude installs the published copy. The npm catalog can differ from this repository.

Without Node, take the directory itself — and turn it into a .zip if an upload field wants one. The tarball is the whole repository (over 100 MB), and the pipe is a POSIX shell recipe (macOS, Linux, WSL) that extracts only this directory:

curl -L https://codeload.github.com/PKU-YuanGroup/OpenAI4S/tar.gz/refs/heads/main \
  | tar -xz --strip-components=2 OpenAI4S-main/skills/using-model-endpoint
python3 -m zipfile -c using-model-endpoint.zip using-model-endpoint

The click-through form of that same download is the repository zip, which is also the route to take on Windows: extract it and copy skills/using-model-endpoint/ out (unzip main.zip 'OpenAI4S-main/skills/using-model-endpoint/*' unpacks only that directory). If you already run OpenAI4S there is nothing to install — the wheel ships every bundled Skill, and a bundled Skill takes precedence over a same-named copy in <data_dir>/user-skills. Targets, provenance, and what the installer refuses to do: tools/skills-installer/.

Files