CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

esm

Comprehensive toolkit for protein language models including ESM3 (generative multimodal protein design across sequence, structure, and function) and ESM C (efficient protein embeddings and representations). Use this skill when working with protein sequences, structures, or function prediction; designing novel proteins; generating protein embeddings; performing inverse folding; or conducting protein engineering tasks. Supports both local model usage and cloud-based Forge API for scalable inference.

Quality

80%

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

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/biology/esm/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

61%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 well-structured, actionable skill body with real, mostly executable ESM SDK code and a clean one-level reference layout. The main weaknesses are recurring redundancy with the reference files (notably the duplicated model tables and install commands) and the absence of explicit validation checkpoints in the generation and batch-processing workflows.

Suggestions

Collapse the Installation section to a single command list plus the flash-attn variant and a pointer to the Forge token URL; the third block repeats 'uv pip install esm' with no new information.

Replace the inline model tables in the Model Selection Guide with a brief selection-criteria list and pointers to the model tables already maintained in references/esm3-api.md and references/esm-c-api.md to cut duplication.

Add explicit validation steps to the workflows: after generation, verify outputs (e.g., check the returned sequence contains no '_' mask tokens or run a structure-prediction pass), and wrap the Forge batch example in error handling/retry so the async gather does not silently drop failures.

DimensionReasoningScore

Conciseness

The body is mostly dense, useful API guidance rather than concept explanation, but it could be tightened: the Installation section repeats "uv pip install esm" three times (the Forge block adds no command), the Model Selection Guide duplicates tables that appear in references/esm3-api.md and references/esm-c-api.md, and several Best Practices bullets ("Implement error handling and retry logic", "Monitor token usage") are generic. This matches the anchor of mostly efficient with some unnecessary content that could be trimmed — not a 2 because there is no tutorial-style padding, and not a 4 because the redundancy is recurring rather than minor.

3 / 5

Actionability

Every core capability ships with a concrete, mostly executable Python snippet using the real ESM SDK (ESM3.from_pretrained, ESMProtein, GenerationConfig, ESM3ForgeInferenceClient, ESMC.from_pretrained), and installation commands are given verbatim. It is not a 5 because of placeholders and gaps: "<token>" must be substituted, "MPRTKEINDAGLIVHSP..." ellipsis sequences are not runnable as-is, and the structure-prediction snippet references a `model` variable defined only in an earlier section.

4 / 5

Workflow Clarity

Multi-step processes are clearly sequenced (chain-of-thought generation as numbered Steps 1–3 across structure, sequence, and function tracks; the async Forge batch executor), but there are no explicit validation or verification checkpoints — generation outputs are never checked and the batch workflow runs with no error handling, which caps this score at 3 per the batch-operation rule. Validation is mentioned only abstractly ("Validate generated sequences with structure prediction or wet-lab experiments" in Best Practices) rather than as a workflow step, so it does not reach a 4.

3 / 5

Progressive Disclosure

The body is a well-organized overview with clearly signaled one-level-deep references — all four cited files (references/esm3-api.md, esm-c-api.md, forge-api.md, workflows.md) exist, are described with what each contains, and are linked at the point of use ("See `references/forge-api.md` for detailed Forge API documentation"). It is not a 5 because some inlined content (model tables, the related-skills list, and capability-level API detail) overlaps with the reference files, leaving minor organization gaps typical of the anchor 4.

4 / 5

Total

14

/

20

Passed

Description

92%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 that explicitly states concrete capabilities for both model families and includes a detailed, explicit trigger clause covering the main use cases. The only weakness is missing natural synonyms and file-format terms users might mention.

DimensionReasoningScore

Specificity

Multiple concrete capabilities are named explicitly — "generative multimodal protein design across sequence, structure, and function", "protein embeddings and representations", "performing inverse folding", "function prediction" — with coverage of both model families and both deployment modes (local and Forge cloud). This matches the anchor for multiple specific concrete actions with comprehensive coverage; it is not a 4 because no meaningful capability area is missing.

5 / 5

Completeness

Both questions are answered explicitly: the "what" is a toolkit for ESM3 generative design and ESM C embeddings, and the "when" is a full "Use this skill when working with protein sequences, structures, or function prediction; designing novel proteins; generating protein embeddings; performing inverse folding; or conducting protein engineering tasks" clause with concrete trigger phrases. This mirrors the anchor-5 example structure; a 4 would require the when-clause to be less explicit or specific.

5 / 5

Trigger Term Quality

Good coverage of natural phrases users would say ("protein sequences", "designing novel proteins", "protein embeddings", "protein engineering tasks"), matching the anchor of good coverage with a few natural terms missing. It is not a 5 because common variations and file formats users actually mention ("amino acid sequence", "structure prediction", .fasta, .pdb, UniProt) are absent, and not a 3 because the included terms are numerous and phrased the way users would say them.

4 / 5

Distinctiveness Conflict Risk

The description occupies a clear niche — the ESM/ESM3/ESM C protein language model families with the Forge API — and its triggers (inverse folding, protein embeddings, ESM3) are distinct from generic structure-prediction or bioinformatics skills. Minor residual overlap exists with "structures" and "function prediction", but it is below the "minor overlap risk with closely related skills" threshold of a 4.

5 / 5

Total

19

/

20

Passed

Validation

87%

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

Validation — 14 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

14

/

16

Passed

Repository

synthetic-sciences/openscience

Path

backend/cli/skills/biology/esm/SKILL.md

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.