CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

clip

OpenAI's model connecting vision and language. Enables zero-shot image classification, image-text matching, and cross-modal retrieval. Trained on 400M image-text pairs. Use for image search, content moderation, or vision-language tasks without fine-tuning. Best for general-purpose image understanding.

Quality

72%

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

Fix and improve this skill with Tessl

tessl review fix ./skills/multimodal/clip/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

57%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-sectioned, code-heavy reference that is genuinely actionable for the common CLIP use cases. Its main weaknesses are marketing-flavored padding (GitHub-star metrics, repeated license info), code fragments that assume state from earlier examples, no validation guidance for batch or moderation workflows, and a references/applications.md bundle file that the body never points to despite duplicating its content.

Suggestions

Drop the "Metrics" section and the duplicate star count/license mentions (they are time-sensitive and already on the linked GitHub repo), and remove obvious comments like "# Load model" to tighten token usage.

Link references/applications.md from the body (e.g., an "## Applications" section pointing to it) and move the per-use-case recipes it duplicates — image search, content moderation, classification variants — out of SKILL.md into that file.

Add validation checkpoints: handle failed or unreadable images in the batch-processing loop, and in content moderation flag results below a confidence threshold instead of always returning the argmax category.

DimensionReasoningScore

Conciseness

The body is mostly lean, code-first prose, but it carries unnecessary padding: a marketing "Metrics" section ("25,300+ GitHub stars", "MIT License") whose star count and license are repeated again in Resources and the frontmatter, obvious comments like "# Load model", and a duplicated label list inside the quick-start example. This matches anchor 3 (mostly efficient but could be tightened) rather than anchor 4's minor-trim level, and the time-sensitive star counts would require upkeep.

3 / 5

Actionability

Installation commands and the zero-shot classification quick start are copy-paste executable, and each use case has concrete code. However, later sections (image-text similarity, content moderation, the Chroma integration) are fragments that silently depend on variables (model, image, text, image_paths, image_embeddings) defined in earlier examples, which is exactly anchor 4's "concrete code with minor gaps" rather than the fully self-contained anchor 5.

4 / 5

Workflow Clarity

Each use case is presented as a short, unambiguous linear recipe, but there are no validation or verification steps anywhere: the batch-processing section does not handle failed image loads or empty batches, and the content-moderation section acts on a single argmax with no confidence threshold or uncertain-class handling. Per the rubric's cap for batch operations without feedback loops, workflow clarity cannot exceed 3.

3 / 5

Progressive Disclosure

The bundle contains references/applications.md, yet the body never mentions or links it — its application recipes are instead duplicated inline (classification, image search), and Resources lists only external URLs. This matches anchor 3's "references present but not clearly signaled; content that should be separate is inline" — structure is decent, but the orphaned reference file and inline duplication prevent anchor 4.

3 / 5

Total

13

/

20

Passed

Description

75%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 solid description with an explicit "Use for..." trigger clause and multiple concrete, natural-sounding capabilities. It is held back from top marks by a generic closing claim ("Best for general-purpose image understanding") and a broad "vision-language tasks" catch-all that slightly blur specificity and distinctiveness.

DimensionReasoningScore

Specificity

Lists several concrete actions — "zero-shot image classification, image-text matching, and cross-modal retrieval" plus "image search, content moderation" — but the closing "Best for general-purpose image understanding" is generic filler and coverage has minor gaps (e.g., embedding computation is never named), matching anchor 4 rather than the comprehensive anchor 5.

4 / 5

Completeness

Clearly answers "what" ("Enables zero-shot image classification, image-text matching, and cross-modal retrieval") and has an explicit "Use for image search, content moderation, or vision-language tasks without fine-tuning" trigger clause, so it exceeds anchor 3's missing-when case; however "vision-language tasks" is a broad catch-all where the when-guidance could be more specific, keeping it at anchor 4 rather than 5.

4 / 5

Trigger Term Quality

Good natural keyword coverage — "image search", "content moderation", "image classification" are phrases users would actually say — but "cross-modal retrieval" is jargon and common variations like "image similarity", "photo search", or "NSFW detection" are missing, so it falls just short of the comprehensive synonym coverage of anchor 5.

4 / 5

Distinctiveness Conflict Risk

The zero-shot / no-fine-tuning / cross-modal-retrieval framing carves a clear niche distinct from captioning or VQA models, but "vision-language tasks" and "general-purpose image understanding" leave minor overlap risk with closely related multimodal skills (BLIP, LLaVA-style), matching anchor 4 rather than the minimal-conflict anchor 5.

4 / 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

frontmatter_unknown_keys

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

Warning

Total

15

/

16

Passed

Repository

OpenLAIR/dr-claw

Path

skills/multimodal/clip/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.