CtrlK
BlogDocsLog inGet started
Tessl Logo

experiment-tracking-swanlab

Provides guidance for experiment tracking with SwanLab. Use when you need open-source run tracking, local or self-hosted dashboards, and lightweight media logging for ML workflows.

71

Quality

87%

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

85%

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

A highly actionable, well-structured reference skill with executable examples across all media types and integrations, and clean one-level-deep progressive disclosure to two real reference files. Its single weakness is conciseness: a full boilerplate PyTorch training loop and some repetition between Quick Start and Core Concepts could be trimmed.

Suggestions

Trim the standard PyTorch training-loop boilerplate in the "With PyTorch" Quick Start to just the swanlab-relevant calls (init, log, finish), or move the full loop into references/integrations.md.

De-duplicate the config-tracking and metric-logging examples between Quick Start and Core Concepts so each concept is shown once.

Consider folding the "Project"/"Experiment" one-line definitions into the first example rather than stating them as separate prose, since Claude already understands these concepts.

DimensionReasoningScore

Conciseness

The body is mostly lean code and concise bullets, but it includes a full standard PyTorch training-loop (optimizer.zero_grad/loss.backward/optimizer.step) Claude already knows, and Core Concepts repeats config and metric-logging examples already shown in Quick Start, so it could be tightened. Not a 3 because not every token earns its place; not a 1 because there is no verbose prose explaining concepts Claude knows.

2 / 3

Actionability

It provides fully executable, copy-paste-ready examples throughout — swanlab.init with config, swanlab.log, every media type (Image, Audio, Text, Video, Object3D, Molecule), echarts, local mode, and framework integrations. Not a 2 because the guidance is concrete and complete rather than pseudocode or incomplete.

3 / 3

Workflow Clarity

The core init → log → finish lifecycle is unambiguous and clearly sequenced in Quick Start, and the local-mode watch/sync flow is similarly clear; there are no destructive or batch operations that would require validation checkpoints, so the simple-skill guidance applies. Not a 2 because the sequence is explicit and unambiguous with no missing risky-operation validation.

3 / 3

Progressive Disclosure

SKILL.md is a clear overview that splits advanced content into two well-signaled, one-level-deep references (references/visualization.md and references/integrations.md, both real files), with a "See Also" section for navigation. Not a 2 because the references are clearly signaled and appropriately separated rather than inline or nested.

3 / 3

Total

11

/

12

Passed

Description

90%

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 well-constructed description with an explicit "Use when" clause, natural trigger terms, and a distinct named-product niche. Its only weakness is the abstract "Provides guidance for" lead, which names capability areas rather than concrete actions, capping specificity at 2.

DimensionReasoningScore

Specificity

The lead phrase "Provides guidance for experiment tracking" is abstract, and while it names capability areas ("open-source run tracking", "local or self-hosted dashboards", "lightweight media logging"), it does not list multiple concrete actions like logging metrics or comparing runs, matching the anchor that names domain and some actions but is not comprehensive. Not a 3 because it lacks a list of specific concrete actions; not a 1 because it does name concrete capability areas rather than being purely vague.

2 / 3

Completeness

It explicitly answers what ("Provides guidance for experiment tracking with SwanLab") and when ("Use when you need open-source run tracking, local or self-hosted dashboards, and lightweight media logging for ML workflows") with an explicit "Use when" trigger clause. Not a 2 because the when-clause is explicit rather than implied.

3 / 3

Trigger Term Quality

It covers natural terms the target audience would say — "experiment tracking", "ML workflows", "run tracking", "media logging", "dashboards", plus the product name "SwanLab" — giving good coverage of natural trigger terms. Not a 2 because the term coverage is broad rather than missing common variations.

3 / 3

Distinctiveness Conflict Risk

The named product "SwanLab" combined with the open-source / self-hosted / media-logging niche gives it a clear, distinct trigger set unlikely to fire for unrelated skills. Not a 2 because the named product and self-hosted angle carve out a specific niche rather than overlapping generically.

3 / 3

Total

11

/

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

frontmatter_unknown_keys

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

Warning

Total

15

/

16

Passed

Repository
Orchestra-Research/AI-Research-SKILLs
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.