CtrlK
BlogDocsLog inGet started
Tessl Logo

chai1

Structure prediction for protein, nucleic-acid, and small-molecule complexes with the Chai-1 foundation model (Chai Discovery 2024, github.com/chaidiscovery/chai-lab). Reach for this skill to predict an antibody-antigen or protein-ligand complex from a single FASTA, to re-fold designed binders as an AlphaFold-multimer alternative, or to drive co-folding from Python for batched campaigns on a GPU.

74

Quality

93%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

87%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.

Concise, highly actionable content anchored by an executable example and sharp operational gotchas (CHAI_DOWNLOADS_DIR, ESM VRAM cost, MSA tradeoff). The main gap is workflow clarity: as a batch job it lacks an explicit validation/checkpoint loop, which caps that dimension.

Suggestions

Add a short numbered run workflow with an explicit validation checkpoint, e.g. 1) ensure output_dir is empty, 2) run inference, 3) check aggregate_score/iptm on outputs before keeping samples.

Turn the error table's recovery actions into a feedback loop (on PermissionError -> set CHAI_DOWNLOADS_DIR and retry; on CUDA OOM -> lower num_diffn_timesteps/split chains and retry) so the batch path has validate->fix->retry semantics.

State the confidence/clash filtering step inline (currently deferred to 'Next') so the keep/discard decision is part of the documented workflow rather than only a pointer.

DimensionReasoningScore

Conciseness

Lean body that assumes competence — no 'what is a protein / what is co-folding' filler — and every section earns its tokens with positioning, runnable code, or a concrete gotcha. The Boltz-2/AlphaFold3 framing is useful differentiation, not generic padding.

5 / 5

Actionability

Copy-paste-ready Python calling run_inference with concrete params, an equivalent shell invocation (chai-lab fold ... --use-msa-server), exact output filenames (pred.model_idx_{0..4}.cif), and a concrete ranking threshold (iptm > 0.5) — covers the common cases fully.

5 / 5

Workflow Clarity

This is a batch operation that writes many files and refuses a non-empty output_dir, yet the guidance notes only 'clear or rotate it between calls' and a soft pass threshold without an explicit validate->fix->retry sequence; per the batch-operation cap, missing feedback-loop checkpoints holds this at 3 rather than 4.

3 / 5

Progressive Disclosure

No bundle files are present and none are needed; the body is self-contained and cleanly sectioned (Running it, three named gotcha sections, an error table, Next pointer), so the simple-skill-with-good-organization bar is met.

5 / 5

Total

18

/

20

Passed

Description

100%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 high-quality description: it names concrete capabilities, surfaces natural trigger terms, answers both 'what' and 'when' explicitly, and carves out a distinct niche against AlphaFold-multimer. Third-person voice is maintained throughout.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'predict an antibody-antigen or protein-ligand complex from a single FASTA', 're-fold designed binders as an AlphaFold-multimer alternative', and 'drive co-folding from Python for batched campaigns on a GPU' — covering the model's use surfaces comprehensively.

5 / 5

Completeness

Explicitly states both what it does ('Structure prediction for protein, nucleic-acid, and small-molecule complexes') and when to reach for it via the 'Reach for this skill to predict... to re-fold... or to drive...' trigger clause.

5 / 5

Trigger Term Quality

Uses natural domain phrasing a user would actually say — 'antibody-antigen', 'protein-ligand complex', 'AlphaFold-multimer alternative', 'co-folding from Python', 'FASTA', 'GPU' — with strong synonym/variant coverage.

5 / 5

Distinctiveness Conflict Risk

Occupies a clear niche (Chai-1 co-folding) and is explicitly contrasted with AlphaFold-multimer, giving it distinct triggers and minimal overlap with generic modeling skills.

5 / 5

Total

20

/

20

Passed

Validation

81%

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

Validation13 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

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
Reviewed

Table of Contents

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.