CtrlK
BlogDocsLog inGet started
Tessl Logo

mle-workflow

Production machine-learning engineering workflow for data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback. Use when building, reviewing, or hardening ML systems beyond one-off notebooks.

67

Quality

81%

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

The canonical home for this skill is mle-workflow in affaan-m/ECC

SKILL.md
Quality
Evals
Security

Quality

Content

62%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, comprehensive MLE workflow with strong sequencing, explicit promotion/rollback validation, and useful copy-paste templates and code. Its main weaknesses are length and conceptual restatement that a competent model already knows, plus a monolithic single-file structure that does not leverage progressive disclosure for a skill of this size.

Suggestions

Tighten 'Metric and Mistake Economics' and 'Data and Feature Hypotheses' to MLE-specific decision heuristics only, dropping generic precision/recall/F1/AUC and missing-value/outlier primers that Claude already knows, to improve conciseness.

Move the large SWE-surface and 'Ten MLE Task Simulations' tables into a references/ file (e.g. references/swe-surface.md and references/task-simulations.md) linked from a concise overview in SKILL.md, so the body acts as an overview with one-level-deep navigation.

Add a second executable code snippet covering the serving/packaging lane (e.g. artifact schema validation or input-schema rejection) so actionability extends across more of the stated workflow rather than only config and promotion gates.

DimensionReasoningScore

Conciseness

The body is long and several sections restate tradeoffs Claude already knows (precision vs recall, F1 vs AUC, missing-value/outlier handling in 'Metric and Mistake Economics' and 'Data and Feature Hypotheses'), plus aphoristic padding such as 'Simplicity is not laziness' and 'The strongest MLE loop is not train -> metric -> ship' — efficient in places but could be tightened.

2 / 3

Actionability

Provides genuinely copy-paste-ready artifacts (Iteration Compact and Observation Ledger templates, the TrainingConfig/artifact_name and PROMOTION_GATES/assert_promotion_ready code, review checklist), but large portions are consideration-lists and principles, and the executable code covers only config and promotion-gating rather than the full workflow breadth.

2 / 3

Workflow Clarity

A 6-step sequenced Core Workflow is backed by explicit validation checkpoints — fail-closed promotion gates, 'do not ship' thresholds, required rollback plans, the Error Analysis feedback loop, and a review checklist — matching the clear-sequence-with-validation anchor.

3 / 3

Progressive Disclosure

Headers and sectioning are clear, but the skill is a single ~340-line monolith with no bundle files (references/scripts/assets absent); the large SWE-surface and 10-task-simulation tables and detailed conceptual sections are inline rather than split into one-level-deep reference files, and the under-50-line exception does not apply.

2 / 3

Total

9

/

12

Passed

Description

100%

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, concrete description that states specific capabilities, provides an explicit 'Use when' trigger with natural keywords, and occupies a clear niche distinct from generic engineering skills. It scores at the top of every dimension with no over-claims or fluff.

DimensionReasoningScore

Specificity

Names multiple concrete capability areas — 'data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback' — rather than vague language, matching the multiple-specific-actions anchor.

3 / 3

Completeness

Explicitly answers both what it does (production ML engineering workflow for the listed capabilities) and when to use it via the 'Use when building, reviewing, or hardening ML systems beyond one-off notebooks' trigger.

3 / 3

Trigger Term Quality

Includes natural phrasing a user would say ('building, reviewing, or hardening ML systems', 'one-off notebooks', 'ML systems') with good coverage of common variations.

3 / 3

Distinctiveness Conflict Risk

Carves a clear niche (production ML systems beyond one-off notebooks) with distinct triggers unlikely to fire for general SWE or data-analysis skills.

3 / 3

Total

12

/

12

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.

Validation15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

Total

15

/

16

Passed

Repository
affaan-m/ECC
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.