alphafold2

Predict protein structure for monomers and multimers with AlphaFold2 via the ColabFold runner (Mirdita et al. 2022, github.com/sokrypton/ColabFold; AlphaFold2 Jumper et al. 2021). Reach for this skill to fold a sequence or complex with the AF2/AF2-Multimer evoformer, to validate designed sequences by self-consistency pLDDT, ipTM, and RMSD, or to run a quick MSA-backed prediction using the public MMseqs2 server.

Quality

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

75%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-quality, operational body: an executable one-command quick start, two genuinely non-obvious gotchas (the gVisor unified-memory hang and the shared MSA server) with concrete mitigations, and a symptom→fix error table. It falls short of top marks only on small refinements — a redundant license paragraph, two non-executable fix hints, and no explicit validate-before-proceeding checkpoint sequence.

Suggestions

Give the unified-memory fix as a full executable invocation (e.g., TF_FORCE_UNIFIED_MEMORY=0 XLA_PYTHON_CLIENT_MEM_FRACTION=0.95 colabfold_batch ...) or an exact sed command for batch.py.

Trim the license-attribution paragraph — it duplicates the frontmatter metadata.third_party block.

Add a brief ordered checklist for interpreting outputs (read rank_001 PDB, check scores JSON plddt/iptm against thresholds, then proceed) to make the validation loop explicit.

Dimension	Reasoning	Score
Conciseness	Dense, non-obvious operational content (gVisor unified-memory loop, MSA-server caching strategy, symptom→fix table) with no basic-concept padding, but the license-attribution sentence duplicates the frontmatter metadata and could be trimmed — anchor 4, not 5.	4 / 5
Actionability	Copy-paste-ready main command with flags, exact env override values, and a concrete --msa-only cache-and-reuse workflow, but the sed-patch fix and the env-override invocation are directional rather than executable commands — anchor 4, not 5.	4 / 5
Workflow Clarity	Clear sequence (run → read rank-1 PDB → gate on ipTM > 0.5), a well-sequenced campaign path, and an error table providing recovery loops, but no explicit ordered step list with validate-before-proceeding checkpoints — anchor 4, not 5.	4 / 5
Progressive Disclosure	No bundle files exist (verified: references/, scripts/, assets/ absent) and all inline content is operational, with well-signaled sections; at ~67 lines the body sits above the simple-skill threshold, so the error table or gVisor gotcha could live one level deep — anchor 4, not 5.	4 / 5
	Total	16 / 20 Passed

Description

83%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 strong description: concrete capabilities, an explicit 'Reach for this skill' trigger clause enumerating three use cases, and consistent third-person voice. It sits just below top marks only on trigger synonym coverage and because closely related folding skills could compete for the same requests.

Suggestions

Add one or two natural user phrasings for the validation use case (e.g., 'check whether a designed sequence folds to its target backbone') so that request triggers the skill reliably.

Mention the output artifacts users ask about by name (ranked PDB models, pLDDT/ipTM scores) to sharpen trigger matching and distinguish from sibling folding skills.

Dimension	Reasoning	Score
Specificity	Multiple concrete actions with named metrics ('Predict protein structure for monomers and multimers', 'fold a sequence or complex', 'validate designed sequences by self-consistency pLDDT, ipTM, and RMSD'), with only minor gaps such as templates/relaxation — anchor 4, not 5.	4 / 5
Completeness	Explicitly answers both 'what' ('Predict protein structure for monomers and multimers with AlphaFold2 via the ColabFold runner') and 'when' with a concrete enumerated trigger clause ('Reach for this skill to fold a sequence or complex..., to validate designed sequences..., or to run a quick MSA-backed prediction...'); third-person voice throughout.	5 / 5
Trigger Term Quality	Good natural-term coverage ('fold a sequence', 'complex', 'AlphaFold2'/'AF2', 'MSA', 'protein structure', 'multimer'), but common phrasings users would say for the validation use case are only partially covered — anchor 4, not 5.	4 / 5
Distinctiveness Conflict Risk	Clear niche (AF2/ColabFold folding) with specific triggers, but sibling folding skills (boltz, chai1, openfold3) satisfy overlapping 'fold this complex' requests, leaving minor overlap risk with closely related skills — anchor 4.	4 / 5
	Total	17 / 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: aipoch/open-science
Path: resources/skills/alphafold2/SKILL.md
Commit: bf35648

Reviewed: 8 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.