ligandmpnn

Inverse-fold a backbone with ligand, nucleic-acid, and metal context using LigandMPNN (Dauparas et al. 2023, github.com/dauparas/LigandMPNN). Reach for this skill to redesign the residues lining a binding pocket around a bound small molecule or cofactor, to design metal-coordinating sites where the geometry must be respected, or to get threaded designed-sequence PDBs out of any MPNN run.

Quality

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

76%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A high-signal, executable skill body that nails the install/run path and error recovery. Its main weakness is workflow clarity: a batch inverse-folding run lacks an explicit validation checkpoint before trusting outputs, even though a ligand-confidence diagnostic is mentioned as a gotcha.

Suggestions

Add an explicit post-run validation step to the "Running it" flow — e.g. after run.py, check that each FASTA header carries a non-zero ligand_confidence and spot-fold one design before scaling batch_size/number_of_batches.

Promote the ligand-confidence check from a gotcha into a numbered checkpoint so the batch workflow reads as: run → verify ligand_confidence > 0 → (only then) scale up.

Consider moving the "Errors worth recognizing" table and the ProDy/ligand-context gotchas into a references/ file (e.g. TROUBLESHOOTING.md) to lift progressive_disclosure and shorten the top-level body.

Dimension	Reasoning	Score
Conciseness	Dense and expert-calibrated with no concept-explaining padding, but a few section headers are sentence-length (e.g. "ProDy compiles from source on py3.11 — pip install fails without a C compiler") and one inline aside could be trimmed.	4 / 5
Actionability	Provides a copy-paste-ready pip/git/run.py invocation with real flags, the sed patch for the numpy break, and an error table mapping each symptom to a concrete fix.	5 / 5
Workflow Clarity	The run sequence is clear and a diagnostic exists ("If ligand_confidence... is missing or zero... do not trust the sequences"), but this batch operation has no explicit validate-then-proceed checkpoint in the run flow, capping the score at 3 per the batch-operation guideline.	3 / 5
Progressive Disclosure	Well-sectioned single-file overview with a one-level "Next:" pointer to boltz/chai1; no bundle files exist to disclose, and the model-type table, gotchas, and error table are appropriately inlined, though the gotcha/error material could justify separate reference files.	4 / 5
	Total	16 / 20 Passed

Description

87%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A tight, third-person description that pairs a concrete capability statement with an explicit "Reach for this skill to..." trigger clause covering three real use cases. It is specific and distinct; only keyword synonym/extension coverage keeps trigger_term_quality and specificity off the top anchor.

Dimension	Reasoning	Score
Specificity	Lists three concrete actions — "redesign the residues lining a binding pocket", "design metal-coordinating sites", and "get threaded designed-sequence PDBs" — but stops short of the comprehensive multi-action coverage of a 5.	4 / 5
Completeness	Clearly states what it does ("Inverse-fold a backbone with ligand, nucleic-acid, and metal context") and explicitly when to use it via "Reach for this skill to..." with three concrete trigger scenarios.	5 / 5
Trigger Term Quality	Natural terms like "binding pocket", "metal-coordinating sites", "redesign the residues", and "threaded designed-sequence PDBs" are present, but common synonyms and file extensions (.pdb, cofactor variants) are not covered.	4 / 5
Distinctiveness Conflict Risk	A clear niche (ligand/NA/metal-aware inverse folding) with triggers distinct from vanilla protein-only MPNNs; minimal overlap risk with sibling skills.	5 / 5
	Total	18 / 20 Passed

Validation

81%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	13 / 16 Passed

Repository: UnicomAI/wanwu
Path: configs/microservice/bff-service/configs/agent-skills/claude-science/ligandmpnn/SKILL.md
Commit: 030c20b

Reviewed: 14 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.