CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

kermt-infer

Run predictions with a finetuned KERMT checkpoint on a SMILES-only CSV. The skill validates that the input ckpt has task FFN heads (refuses pretrain ckpts with a redirect to kermt-finetune), validates the CSV, prepares the data (clean + rdkit_2d features), then launches main.py predict inside the kermt container (blocking, minutes-scale).

Quality

81%

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

88%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.

The body is a tightly organized, fully actionable workflow with strong validation checkpoints and error-recovery guidance; its only weak spots are minor repetition and a referenced config/ directory that is absent from the bundle.

Suggestions

Remove the duplicated 'Multi-GPU inference is not currently supported' line, keeping it only in one section (e.g. Hardware requirements or Hard rules).

Either add the referenced config/ directory to the bundle or correct the 'defaults are bundled in config/' statement to point at the real location of the defaults.

DimensionReasoningScore

Conciseness

Lean and well-sectioned with no explanations of concepts Claude already knows; the only trimming opportunity is the repeated 'Multi-GPU inference is not currently supported' line appearing in three sections, which is minor padding.

4 / 5

Actionability

Provides fully executable, copy-paste-ready commands with absolute container paths and real script names, plus a Common errors section mapping real error strings to fixes — covering the common cases concretely.

5 / 5

Workflow Clarity

Seven steps are explicitly sequenced with validation checkpoints ('Refuse to proceed on ok: false', 'Abort on ok: false') and a Common errors feedback loop, satisfying the rubric's requirement for explicit validation in batch operations.

5 / 5

Progressive Disclosure

Good section structure with one-level-deep references to real bundled scripts via $SKILL_DIR/scripts/, but the body states 'defaults are bundled in config/' while no config/ directory exists in the bundle — a minor dangling reference gap.

4 / 5

Total

18

/

20

Passed

Description

67%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.

The description is concrete and highly distinctive, naming the exact workflow steps and constraints, but it is purely a 'what' statement with no 'when to use' trigger guidance and relies on domain jargon over natural user phrasing.

Suggestions

Add an explicit 'Use when...' clause naming natural trigger phrases (e.g. 'Use when the user wants to run inference / get predictions from a finetuned KERMT model on a SMILES CSV').

Include natural synonyms and file extensions users might say (e.g. '.csv', 'inference', 'predict on molecules') alongside the domain terms.

Soften jargon where a natural phrasing would do (e.g. pair 'task FFN heads' with 'finetuned for a prediction task').

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'validates that the input ckpt has task FFN heads', 'validates the CSV', 'prepares the data (clean + rdkit_2d features)', 'launches main.py predict' — giving comprehensive coverage of the workflow rather than vague abstractions.

5 / 5

Completeness

Clearly answers 'what' (validate ckpt/CSV, prepare data, run predict) but provides no 'Use when...' or equivalent trigger guidance, which per the rubric caps completeness at 3.

3 / 5

Trigger Term Quality

Contains relevant domain terms ('predictions', 'checkpoint', 'SMILES-only CSV', 'finetuned') but leans on technical jargon (KERMT, FFN heads, rdkit_2d) and omits common natural synonyms or file extensions a user might actually say.

3 / 5

Distinctiveness Conflict Risk

Occupies a clear niche — finetuned KERMT inference on SMILES-only CSVs with FFN-head validation — with distinct, highly specific triggers and minimal risk of firing for an unrelated skill.

5 / 5

Total

16

/

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

NVIDIA/skills

Path

skills/bionemo-kermt-infer/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.