aiskillstore/marketplace / skills/abdulsamad94/rag-pipeline
rag-pipeline
VerifiedDetails on the Retrieval Augmented Generation pipeline, Ingestion, and Vector Search.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 1.0 KB |
| skill-report.json | config | 16.8 KB |
skill-report.json
View on GitHub{
"schema_version": "2.0",
"meta": {
"generated_at": "2026-07-18T09:53:09.848Z",
"slug": "abdulsamad94-rag-pipeline",
"source_url": "https://github.com/AbdulSamad94/Hackhaton-SpecsKitPlus/tree/master/.claude/skills/rag",
"source_ref": "master",
"model": "claude",
"analysis_version": "3.0.0",
"source_type": "community",
"content_hash": "f7dc70ef14db8e99127f7f084cbd298e80c9aa1741c46a64c02c6941eab9d1cd",
"tree_hash": "53b088d85fd149a3826a2210ec919af881ff84a82a69ae8a6a99073075ac676b",
"provenance": {
"model_effective": "claude",
"fallback_chain": [
{
"agent": "codex",
"outcome": "failed",
"error": "2026-07-18T09:53:09.980569Z ERROR codex_models_manager::manager: failed to refresh available models: Missing environment variable: `HELM_API_KEY`.\n2026-07-18T09:53:10.003199Z ERROR codex_models_manage…"
},
{
"agent": "claude",
"outcome": "success"
}
],
"analysis_version": "3.0.0"
}
},
"skill": {
"name": "RAG Pipeline",
"description": "Reference documentation describing a Retrieval Augmented Generation pipeline, covering document ingestion, embedding, and vector search with Qdrant.",
"summary": "Documentation for a RAG pipeline covering ingestion, embedding, and Qdrant vector search.",
"icon": "📦",
"author": "AbdulSamad94",
"license": "MIT",
"category": "documentation",
"tags": [
"rag",
"vector-search",
"qdrant",
"embeddings",
"documentation"
],
"supported_tools": [
"claude",
"codex",
"claude-code"
],
"risk_factors": [
"external_commands"
]
},
"security_audit": {
"risk_level": "safe",
"is_blocked": false,
"safe_to_publish": true,
"analysis_status": "ok",
"summary": "This skill is a single Markdown documentation file (SKILL.md) that explains the architecture of a Retrieval Augmented Generation pipeline. All 10 static findings are false positives: the pattern matcher flagged Markdown inline code spans (backtick-delimited text) as Ruby/shell backtick execution. There is no executable code, no network calls, and no command execution in the file.",
"remediation": [],
"finding_verdicts": [
{
"id": "external_commands:SKILL.md:9:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `backend/ingest.py` are Markdown inline-code formatting inside a documentation bullet, not shell or Ruby backtick command execution. SKILL.md is a prose Markdown file with no interpreter."
},
{
"id": "external_commands:SKILL.md:11:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `docs/` are Markdown inline-code formatting describing a directory path, not command execution. No shell or Ruby runtime is involved."
},
{
"id": "external_commands:SKILL.md:14:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `models/text-embedding-004` are Markdown inline-code formatting naming an embedding model, not backtick command execution."
},
{
"id": "external_commands:SKILL.md:15:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `physical_ai_book` are Markdown inline-code formatting naming a Qdrant collection, not backtick command execution."
},
{
"id": "external_commands:SKILL.md:16:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `python backend/ingest.py` are Markdown inline-code documenting how a human would run the script; this is descriptive prose, not an execution directive to the agent."
},
{
"id": "external_commands:SKILL.md:19:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `qdrant-client` are Markdown inline-code naming a client library, not backtick command execution."
},
{
"id": "external_commands:SKILL.md:20:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `physical_ai_book` are Markdown inline-code naming a collection, not backtick command execution."
},
{
"id": "external_commands:SKILL.md:25:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `backend/utils/helpers.py` are Markdown inline-code formatting a file path reference, not backtick command execution."
},
{
"id": "external_commands:SKILL.md:27:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `backend/personalization.py` and `software_background` are Markdown inline-code formatting file paths and variable names, not backtick command execution."
},
{
"id": "external_commands:SKILL.md:30:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.95,
"reason": "The backticks around `Agent` and `backend/agents.py` are Markdown inline-code formatting a class name and file path, not backtick command execution."
}
],
"semantic_findings": [],
"critical_findings": [],
"high_findings": [],
"medium_findings": [],
"low_findings": [],
"dangerous_patterns": [],
"files_scanned": 1,
"total_lines": 31,
"audit_model": "claude",
"audited_at": "2026-07-18T09:53:09.848Z",
"static_findings": [
{
"id": "external_commands:SKILL.md:9:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 9,
"line_end": 9,
"snippet": "- **Script**: `backend/ingest.py`"
},
{
"id": "external_commands:SKILL.md:11:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 11,
"line_end": 11,
"snippet": "1. Scans `docs/`."
},
{
"id": "external_commands:SKILL.md:14:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 14,
"line_end": 14,
"snippet": "4. Embeds using `models/text-embedding-004`."
},
{
"id": "external_commands:SKILL.md:15:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 15,
"line_end": 15,
"snippet": "5. Upserts to Qdrant collection `physical_ai_book`."
},
{
"id": "external_commands:SKILL.md:16:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 16,
"line_end": 16,
"snippet": "- **Run**: `python backend/ingest.py`"
},
{
"id": "external_commands:SKILL.md:19:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 19,
"line_end": 19,
"snippet": "- **Client**: `qdrant-client`"
},
{
"id": "external_commands:SKILL.md:20:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 20,
"line_end": 20,
"snippet": "- **Collection**: `physical_ai_book`"
},
{
"id": "external_commands:SKILL.md:25:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 25,
"line_end": 25,
"snippet": "- **File**: `backend/utils/helpers.py`."
},
{
"id": "external_commands:SKILL.md:27:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 27,
"line_end": 27,
"snippet": "- **Personalization**: `backend/personalization.py` creates system instructions based on `software_b"
},
{
"id": "external_commands:SKILL.md:30:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 30,
"line_end": 30,
"snippet": "We use a custom `Agent` class (`backend/agents.py`) that wraps the LLM calls, allowing for future ex"
}
],
"risk_factor_evidence": [
{
"factor": "external_commands",
"evidence": [
{
"file": "SKILL.md",
"line_start": 9,
"line_end": 9
},
{
"file": "SKILL.md",
"line_start": 11,
"line_end": 11
},
{
"file": "SKILL.md",
"line_start": 14,
"line_end": 14
},
{
"file": "SKILL.md",
"line_start": 15,
"line_end": 15
},
{
"file": "SKILL.md",
"line_start": 16,
"line_end": 16
},
{
"file": "SKILL.md",
"line_start": 19,
"line_end": 19
},
{
"file": "SKILL.md",
"line_start": 20,
"line_end": 20
},
{
"file": "SKILL.md",
"line_start": 25,
"line_end": 25
},
{
"file": "SKILL.md",
"line_start": 27,
"line_end": 27
},
{
"file": "SKILL.md",
"line_start": 30,
"line_end": 30
}
]
}
],
"agent_auto_install_policy": "allowed",
"manual_install_policy": "allowed"
},
"content": {
"user_title": "Document a RAG Pipeline Architecture",
"value_statement": "Understanding how a Retrieval Augmented Generation system moves from raw documents to answerable context is hard to track across scattered files. This skill provides a clear reference for the ingestion, embedding, and Qdrant vector search stages of a RAG pipeline.",
"seo_keywords": [
"RAG pipeline",
"retrieval augmented generation",
"Qdrant vector search",
"document ingestion",
"text embeddings",
"Claude",
"Codex",
"Claude Code",
"vector database",
"chunking"
],
"actual_capabilities": [
"Describes the document ingestion process, including scanning, MDX cleaning, and text chunking with defined size and overlap.",
"Explains the embedding step using a named text embedding model.",
"Documents the Qdrant vector search setup, including collection name, vector size, and cosine similarity.",
"Outlines prompt engineering and personalization based on user background attributes.",
"Summarizes an agentic flow that wraps LLM calls in a custom Agent class."
],
"limitations": [
"This is documentation only; it contains no runnable code or scripts.",
"It describes a specific project setup and does not provide a general-purpose RAG framework.",
"It assumes the reader already has the referenced Python files and dependencies in place.",
"It does not include configuration steps for API keys or Qdrant deployment."
],
"use_cases": [
{
"title": "Onboard a new engineer",
"description": "A developer joining the project reads this skill to learn how the RAG pipeline works before touching the code.",
"target_user": "Software engineers new to the codebase"
},
{
"title": "Review the retrieval architecture",
"description": "A technical lead uses the documentation to review chunking, embedding, and vector search choices.",
"target_user": "Technical leads and architects"
},
{
"title": "Explain the pipeline to stakeholders",
"description": "A product owner references the summary to understand how the system retrieves and personalizes context.",
"target_user": "Product owners and non-technical stakeholders"
}
],
"prompt_templates": [
{
"title": "Summarize the pipeline",
"prompt": "Summarize the stages of the RAG pipeline described in this skill.",
"scenario": "A quick overview for a beginner."
},
{
"title": "Explain the ingestion step",
"prompt": "Explain how documents are ingested, cleaned, and chunked in this pipeline.",
"scenario": "Understanding the data preparation flow."
},
{
"title": "Compare the vector search settings",
"prompt": "Describe the Qdrant collection configuration and explain why cosine similarity fits this use case.",
"scenario": "Reviewing the vector search design decisions."
},
{
"title": "Extend the personalization logic",
"prompt": "Based on the personalization section, propose how the system could incorporate additional user background attributes into the system instructions.",
"scenario": "An advanced task extending the described architecture."
}
],
"output_examples": [
{
"input": "Summarize the stages of the RAG pipeline described in this skill.",
"output": "The pipeline has three main stages. First, ingestion scans a documents folder, cleans the text, splits it into chunks, embeds the chunks, and stores them in a Qdrant collection. Second, vector search retrieves relevant chunks using cosine similarity. Third, prompt engineering builds a context-rich prompt and personalizes system instructions based on the user background."
},
{
"input": "Explain how documents are ingested, cleaned, and chunked in this pipeline.",
"output": "During ingestion the system scans the documents folder, removes MDX frontmatter and import statements, and splits the remaining text into chunks of one thousand characters with a one hundred character overlap. Each chunk is then embedded and upserted into the vector store."
}
],
"best_practices": [
"Read the ingestion section first, since later stages depend on how chunks are prepared.",
"Verify that the referenced Python files exist in the codebase before relying on the described behavior.",
"Keep the collection name and vector size consistent between ingestion and search."
],
"anti_patterns": [
"Treating this documentation as executable code that will run on its own.",
"Changing the chunk size or embedding model without updating both ingestion and search.",
"Assuming the pipeline handles API keys or Qdrant deployment, which are out of scope here."
],
"faq": [
{
"question": "What does this skill contain?",
"answer": "It contains reference documentation describing the architecture of a Retrieval Augmented Generation pipeline."
},
{
"question": "Does this skill run any code?",
"answer": "No. It is a Markdown documentation file and does not execute any code or scripts."
},
{
"question": "Which vector database does the pipeline use?",
"answer": "The pipeline uses Qdrant, storing vectors in a collection named physical_ai_book."
},
{
"question": "What are the chunking settings?",
"answer": "Text is chunked into segments of one thousand characters with a one hundred character overlap."
},
{
"question": "How is retrieval personalized?",
"answer": "System instructions are tailored using the user software background and hardware background attributes."
},
{
"question": "What similarity metric does the vector search use?",
"answer": "The vector search uses cosine similarity with a vector size of seven hundred sixty eight."
}
]
},
"file_structure": [
{
"name": "SKILL.md",
"type": "file",
"path": "SKILL.md",
"lines": 31
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |