CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

rag-implementation

Build Retrieval-Augmented Generation (RAG) systems for LLM applications with vector databases and semantic search. Use when implementing knowledge-grounded AI, building document Q&A systems, or integrating LLMs with external knowledge bases.

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

SKILL.md
Quality
Evals
Security

Quality

Content

65%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 delivers a solid executable Quick Start and exemplary one-level-deep progressive disclosure, but it spends tokens on padding (intro sentence, "Purpose" glosses, a When-to-Use list duplicating the description) and omits explicit validation checkpoints in the RAG workflow.

Suggestions

Cut the 'Master Retrieval-Augmented Generation...' intro sentence and the four 'Purpose:' lines — they explain concepts (what vector databases, embeddings, and reranking are) that Claude already knows, which is the conciseness rubric's main penalty.

Add explicit validation checkpoints to the Quick Start workflow: check that retrieval returned documents before generating, and on empty/weak retrieval either rephrase the query or have the model state it cannot answer — currently the only safeguard is implicit inside the prompt template (workflow_clarity).

Trim the 'When to Use This Skill' section, which restates the frontmatter description's triggers, and move the time-sensitive 'Models (2026)' embedding table into references/ (or a dated/deprecated section) so stale model listings don't clutter the always-loaded tier (conciseness).

DimensionReasoningScore

Conciseness

Mostly efficient lists and tables, but it includes unnecessary explanation Claude does not need: the padded intro "Master Retrieval-Augmented Generation (RAG) to build LLM applications that provide accurate, grounded responses", four "Purpose:" lines explaining well-known concepts (e.g., "Convert text to numerical vectors for similarity search"), and a "When to Use This Skill" section that duplicates the frontmatter description's triggers. The time-sensitive "Models (2026)" table is not placed in a deprecated/old-patterns section, which the guidelines penalize. This fits anchor 3 ('mostly efficient but includes some unnecessary explanation') better than anchor 4, where over-explanation would be only a minor instance.

3 / 5

Actionability

The Quick Start is a complete, executable LangGraph example — imports, typed state, retriever setup, prompt, nodes, graph assembly, and invocation through `await rag_chain.ainvoke(...)` — matching 'mostly executable guidance; concrete code with minor gaps'. It is not 5 because the Core Components sections are descriptive catalogs rather than instructions (e.g., "Pinecone: Managed, scalable, serverless") and key operations like chunking are only available via the external reference.

4 / 5

Workflow Clarity

A sequence exists but only implicitly through the graph code (START → retrieve → generate → END); there are no explicit validation checkpoints such as handling empty retrieval or assessing context sufficiency — only the implicit "If you cannot answer, say so" embedded in the prompt template. This matches anchor 3 ('sequence present but checkpoints missing or implicit'); it is not 4 because most checkpoints are absent rather than present with minor gaps.

3 / 5

Progressive Disclosure

The body is a clear overview (component catalogs, quick start) with a dedicated section signaling the one-level-deep reference: "Detailed pattern documentation lives in `references/details.md`. Read that file when the navigation tier above is insufficient." The referenced file exists in the bundle and itself contains no nested references, so content is appropriately split and navigation is easy — matching the anchor 5 example of a well-signaled, one-level-deep reference structure.

5 / 5

Total

15

/

20

Passed

Description

87%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 with the explicit what-plus-when structure, natural trigger terms, and a distinct RAG niche. The only weaknesses are minor: a few natural keywords (embeddings, hallucination) are absent and the capability enumeration could be slightly more comprehensive.

DimensionReasoningScore

Specificity

"Build Retrieval-Augmented Generation (RAG) systems for LLM applications with vector databases and semantic search" names the domain plus several concrete capabilities (vector databases, semantic search, document Q&A, knowledge-base integration), matching the 'several specific actions; minor gaps in coverage' anchor. It is above anchor 3 (only 1-2 actions) but below anchor 5 because capabilities like chunking, reranking, and citation are not mentioned.

4 / 5

Completeness

It clearly answers both questions: the what is "Build Retrieval-Augmented Generation (RAG) systems for LLM applications with vector databases and semantic search" and the when is an explicit "Use when implementing knowledge-grounded AI, building document Q&A systems, or integrating LLMs with external knowledge bases" clause with concrete trigger phrases — the exact structure of the anchor 5 example. It cannot be 4 because the 'when' is already explicit and specific, not merely present.

5 / 5

Trigger Term Quality

Natural terms users would actually say are well covered: "RAG", "vector databases", "semantic search", "document Q&A systems", "external knowledge bases". A few common variations are missing (e.g., "embeddings", "chatbot", "hallucination", specific vector DB names), so it fits 'good keyword coverage; a few natural terms missing' rather than the comprehensive synonym coverage of anchor 5.

4 / 5

Distinctiveness Conflict Risk

"RAG" plus "vector databases", "semantic search", and "document Q&A" define a clear niche with distinct triggers and minimal conflict risk against other plausible skills. It is not 4 because the triggers named are specific enough that overlap with a neighboring skill (e.g., a generic embeddings or chatbot skill) is minimal.

5 / 5

Total

18

/

20

Passed

Validation

100%

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

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

wshobson/agents

Path

plugins/llm-application-dev/skills/rag-implementation/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.