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

71%

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

Fix and improve this skill with Tessl

tessl review fix ./skills/rag-implementation/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

50%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 rich with concrete, copy-paste-style code and good section organization, but it is monolithic, padded with capability catalogs Claude already knows, and several code paths rely on undefined helpers and a deprecated LangChain API. Splitting reference material into bundle files and adding an explicit evaluate-and-iterate checkpoint would raise quality substantially.

Suggestions

Split the advanced patterns, vector store configurations, chunking strategies, and prompt templates into references/*.md files and keep only the Quick Start and decision guidance in SKILL.md.

Update all imports to the current LangChain package layout (langchain_community, langchain_openai) and replace the evaluation snippet's undefined helpers with working implementations (e.g., Ragas metrics or a simple LLM-graded groundedness check).

Turn the Instructions workflow into an explicit feedback loop: after evaluation, give thresholds and remediation steps (e.g., 'if retrieval quality < 0.7, reduce chunk size or add reranking, then re-evaluate').

DimensionReasoningScore

Conciseness

Core code sections are efficient, but the intro restates the description, "Purpose" lines and one-line catalogs of vector stores/embedding models explain concepts Claude already knows, and three near-duplicate prompt templates pad the token budget.

3 / 5

Actionability

Most snippets are concrete, but the evaluation block calls undefined helpers (calculate_accuracy, evaluate_retrieved_docs, check_groundedness), Pattern 3 uses an undefined llm variable, and all imports use the deprecated pre-0.1 langchain paths that break on current versions — gaps beyond 'minor'.

3 / 5

Workflow Clarity

The 4-step Instructions sequence is ordered and coherent, but the evaluation step is an implicit checkpoint with no thresholds or feedback loops (e.g., 'if retrieval quality is low, tune chunk size or add reranking, then re-evaluate').

3 / 5

Progressive Disclosure

Headers and sections are well organized, but ~410 lines live in one monolithic file with no bundle files; vector store configurations, chunking strategies, and prompt templates clearly belong in separate reference files.

3 / 5

Total

12

/

20

Passed

Description

78%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 that explicitly answers both what the skill does and when to use it, with mostly natural trigger terms and a clear niche. The main gaps are limited action coverage and a few missing synonyms users might naturally say.

Suggestions

Add 1-2 more concrete actions to the what-clause (e.g., 'chunk documents, rerank retrieval results, and evaluate groundedness') to lift specificity.

Include missing natural synonyms in the trigger clause such as 'embeddings', 'retrieval pipeline', or 'reducing hallucinations with grounded answers'.

Trim the broad 'for LLM applications' framing to keep triggers scoped tightly to retrieval/knowledge-grounding use cases.

DimensionReasoningScore

Specificity

Names the domain plus 1-2 concrete actions ("Build ... systems", "with vector databases and semantic search") but omits other key RAG actions like chunking, reranking, and evaluation, matching the anchor for 1-2 concrete actions without comprehensive coverage.

3 / 5

Completeness

Clearly states what the skill does ("Build Retrieval-Augmented Generation (RAG) systems ... with vector databases and semantic search") and an explicit "Use when" clause with three concrete trigger scenarios, mirroring the anchor-5 example structure.

5 / 5

Trigger Term Quality

Natural trigger phrases like "vector databases", "semantic search", "document Q&A systems", and "knowledge-grounded AI" are present, but common synonyms such as "embeddings", "retrieval pipeline", or "hallucination reduction" are missing.

4 / 5

Distinctiveness Conflict Risk

RAG is a clear niche with distinct triggers (vector databases, document Q&A, knowledge-grounded AI), but the broad phrase "for LLM applications" creates minor overlap risk with generic LLM or embeddings skills.

4 / 5

Total

16

/

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

rmyndharis/antigravity-skills

Path

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.