CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

foldseek-structural-search

Performs 3D structural searches of proteins against various databases (PDB, AlphaFold, CATH, MGnify, etc.) using the Foldseek API. Use ONLY when the user provides a physical 3D coordinate file (.cif, .mmcif, or .pdb) and wants to find structurally similar proteins. Do NOT use if the user only provides a protein sequence, gene name, or UniProt ID.

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

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 well-structured, actionable skill body: the workflow is clearly sequenced with explicit input/database validation and concrete commands matching the bundled script's allowlist. Main issues are mild redundancy between Core Rules and the Instructions, a malformed trailing error-handling heading, and an unreferenced citation.bib in the bundle.

Suggestions

Merge the redundant Core Rules ('File Requirement', 'Strict Validation') into Instructions steps 1-2, or reduce them to one-line pointers, to remove duplication and tighten the body.

Replace the malformed trailing heading with a proper '## Error Handling' section covering what to do when the API returns an error or the file is missing, including the ask-the-user-to-verify step.

Reference references/citation.bib from the body (e.g., cite the Foldseek paper when reporting results) or remove it from the bundle so all bundle files are discoverable.

DimensionReasoningScore

Conciseness

The body is largely operational with useful domain-specific guidance (metric interpretation is task-relevant, not generic padding), but the Core Rules section duplicates Instructions steps 1-2 ("File Requirement"/"Strict Validation" restate the input and database validation), and "Do Not Parse the JSON" overlaps "No Raw Parsing". This fits anchor 4 (efficient with minor instances that could be trimmed) better than anchor 3, since the redundancy is brief rather than pervasive over-explanation.

4 / 5

Actionability

Concrete executable commands are given for both cases ("uv run scripts/search.py <path-to-file> -o <generated-filename.json> > <generated-filename.md>" and the --databases variant), with a concrete database allowlist and example output filenames. It falls short of anchor 5 because no fully worked example with real values is provided and the error-handling guidance is malformed, leaving minor gaps that fit anchor 4.

4 / 5

Workflow Clarity

Nine steps are clearly sequenced (validate input → validate databases → name outputs → execute → read results → interpret → functional analysis → report files) with explicit halt-on-invalid-input checkpoints in steps 1-2. It does not reach anchor 5 because error recovery is malformed — the final section is a broken heading "## * If the API returns an error or the file is missing, inform the user clearly" followed by an orphaned line "and ask them to verify the file path" — leaving a minor validation/feedback gap characteristic of anchor 4.

4 / 5

Progressive Disclosure

Good structure with clear section headers and a referenced script (scripts/search.py) that exists in the bundle; the ~90-line body appropriately keeps the core workflow inline. Minor gaps keep it at anchor 4 rather than 5: references/citation.bib exists in the bundle but is never referenced in the body, and the dangling final pseudo-heading disrupts navigation.

4 / 5

Total

16

/

20

Passed

Description

90%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: it states a concrete capability, gives explicit and bidirectional trigger guidance, and is clearly distinguishable from sequence-search alternatives. The only weakness is that it names just one action, leaving the skill's reporting and file-saving outputs unmentioned.

DimensionReasoningScore

Specificity

"Performs 3D structural searches of proteins against various databases (PDB, AlphaFold, CATH, MGnify, etc.) using the Foldseek API" names the domain and one concrete action with rich detail (databases, API), but lists no further actions — reporting metrics, functional analysis, and saving results from the body are unmentioned, so coverage is not comprehensive. It clearly matches anchor 3 (1-2 concrete actions, not comprehensive) rather than anchor 4, which requires several distinct listed actions.

3 / 5

Completeness

The "what" is explicit ("Performs 3D structural searches of proteins against various databases... using the Foldseek API") and the "when" is explicit with concrete trigger phrases in both directions: "Use ONLY when the user provides a physical 3D coordinate file (.cif, .mmcif, or .pdb) and wants to find structurally similar proteins. Do NOT use if the user only provides a protein sequence, gene name, or UniProt ID." This clearly matches anchor 5; it exceeds anchor 4 because the trigger conditions are concrete, not merely present.

5 / 5

Trigger Term Quality

The description covers natural terms comprehensively including the tool name ("Foldseek"), natural phrases ("3D structural searches", "find structurally similar proteins"), file extensions (.cif, .mmcif, .pdb), database names (PDB, AlphaFold, CATH, MGnify), and likely-confusable inputs ("protein sequence, gene name, or UniProt ID"). This matches anchor 5's requirement for comprehensive coverage including synonyms and file extensions; only marginal synonyms like "fold" or "structural homolog" are absent.

5 / 5

Distinctiveness Conflict Risk

It occupies a clear niche (structure-file-only search via Foldseek) with an explicit negative trigger ("Do NOT use if the user only provides a protein sequence, gene name, or UniProt ID") that disambiguates it from sequence-homology search skills. Minimal conflict risk, matching anchor 5.

5 / 5

Total

18

/

20

Passed

Validation

93%

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

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

Total

15

/

16

Passed

Repository

google-deepmind/science-skills

Path

skills/foldseek_structural_search/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.