Protein Mutation Enhancement Skill

An iterative protein mutation campaign: build a library, score it, rank it, decide whether to run another round. This is the orchestration layer, not a model. Library enumeration, score merging, ranking and loop control are deterministic and pure stdlib; the ESM, folding and assay numbers come from elsewhere — fair-esm2 for sequence effect, esmfold2 for structure — and are joined in on a stable variant ID like A12V+G47D. A variant ranking first is not evidence of gain of function.

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 protein-mutation-enhancement --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 protein-mutation-enhancement --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/protein-mutation-enhancement
python3 -m zipfile -c protein-mutation-enhancement.zip protein-mutation-enhancement

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/protein-mutation-enhancement/ out (unzip main.zip 'OpenAI4S-main/skills/protein-mutation-enhancement/*' 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

File Responsibility

SKILL.md The input contracts, how single, double and higher-order libraries are built, where the external scores come from (fair-esm2 for sequence effect, esmfold2 for structure), thresholded ranking, the rule for continuing a round or stopping, the practical defaults, and the validation the output still needs.

kernel.py Optional sidecar, pure stdlib. Validates sequences and A12V-style mutation notation, normalizes and applies variants, and enumerates a library deterministically with stable position-sorted IDs, so a score table can safely join on id. It scores the substitution itself with a local heuristic over amino-acid class, hydropathy, charge and volume; reads score tables and writes the library out as FASTA; merges and normalizes the metrics; ranks by weighted composite score; runs a selection round against acceptance thresholds and reports whether to continue; suggests the positions worth opening next; and persists the ranked result as JSON.

File	Responsibility
`SKILL.md`	The input contracts, how single, double and higher-order libraries are built, where the external scores come from (`fair-esm2` for sequence effect, `esmfold2` for structure), thresholded ranking, the rule for continuing a round or stopping, the practical defaults, and the validation the output still needs.
`kernel.py`	Optional sidecar, pure stdlib. Validates sequences and `A12V`-style mutation notation, normalizes and applies variants, and enumerates a library deterministically with stable position-sorted IDs, so a score table can safely join on `id`. It scores the substitution itself with a local heuristic over amino-acid class, hydropathy, charge and volume; reads score tables and writes the library out as FASTA; merges and normalizes the metrics; ranks by weighted composite score; runs a selection round against acceptance thresholds and reports whether to continue; suggests the positions worth opening next; and persists the ranked result as JSON.

The built-in property score is a heuristic term in the composite, not a functional predictor. Whatever comes out on top still needs independent computational and experimental validation.