What is internal-knowledge-base CLAUDE.md?

A project instruction file for building an internal knowledge base that uses document search to answer questions with evidence, then exposes it through an MCP server. MCP is a standard way for an AI tool to connect to an external service.

It guides development of the document-search pipeline and MCP server, including source-format decisions, evaluation reviews, and troubleshooting.

It keeps the build plan, technical decisions, and evaluation process in one place. It also requires detailed analysis of failures instead of relying only on overall test scores.

How kimsb2429/internal-knowledge-base configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to kimsb2429/internal-knowledge-base. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/kimsb2429/internal-knowledge-base/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/kimsb2429/internal-knowledge-base

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/kimsb2429/internal-knowledge-base/claude-md/github.svg)](https://agentmods.dev/instructions/kimsb2429/internal-knowledge-base/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/kimsb2429/internal-knowledge-base/claude-md"><img src="https://agentmods.dev/badge/instructions/kimsb2429/internal-knowledge-base/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/instructions/kimsb2429/internal-knowledge-base/claude-md"><img src="https://agentmods.dev/badge/instructions/kimsb2429/internal-knowledge-base/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 821 This file is loaded in full into every session.

When invoked 821 The same file — it is already loaded in full.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00821	$0.00821
Opus 5.5	$0.00328	$0.00328
Sonnet 5.5	$0.00164	$0.00164
Haiku 4.5	$0.00082	$0.00082

Measured 2mo ago against content hash 715d7302ea82, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

internal-knowledge-base CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 36 lines

How it starts

The opening of the file, as written. The whole thing — 36 lines — stays where its author put it; the contents beside it link to each section on GitHub.

internal-knowledge-base

VA Education RAG pipeline — from raw corpus to MCP server.

Zero-to-MCP plan

The canonical build plan is the "From Zero to Knowledge MCP" section in docs/2026-04-11-engineering-rag-evidence-and-howtos.md. The plan doc is authoritative for any "zero to mcp" references.

Standing rules

Post-eval analysis protocol (apply after every eval run — baseline, Loop A iteration, Loop B iteration, anywhere): never report aggregate metrics alone. Always follow with (a) per-query_type segmentation, (b) 2-3 representative failures pulled per failure cluster with full chunk/judge reasoning, (c) named failure patterns with concrete mechanisms, (d) mapping from each pattern to a specific plan step / lever. The plan doc (§ "From Zero to Knowledge MCP", step 9) has the full template. Anti-pattern: proposing the next intervention directly from aggregate deltas — that tunes blindly.

Key decisions

HTML articles stay as HTML — no markdown conversion. Markdown is lossy for tables (colspan/rowspan). LLMs read HTML fine. One code path for all 237 HTML articles.
Markdown only for VADIR ICD PDF — no HTML source, so parsed markdown + MarkdownHeaderTextSplitter is the right path.
Generic chunker + source-specific preprocessors — chunk_documents.py is source-agnostic; each source gets its own <source>_preprocess.py that normalizes HTML quirks (headings, layout tables, div wrappers) before chunking. Adding a new source means writing a preprocessor, not forking the chunker.
Custom HTML splitter — LangChain's HTMLHeaderTextSplitter strips HTML tags, destroying table structure. We use our own (split_html_by_headings + split_chunk_by_elements in chunk_documents.py) that preserves raw HTML.
Row-group splitting for tables >50K tokens — above 50K, summary indexing breaks down (full content can't be served in a 200K context window). Split into row-groups with repeated headers; each chunk is self-contained.
Heading-only embed_text for oversized chunks — instead of LLM-generated summaries ($13 in Sonnet calls), use title + heading_path for the embed vector. Full content still served at generation. If eval shows content-level queries failing, add a retrieve_full_doc MCP method or upgrade to LLM summaries in Loop A.
Schema: embed_text holds what was embedded (content for regular chunks, heading-only for oversized); content always holds the full original.

Read the full file on GitHub · 36 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 36 lines · 821 tokens per session scan A 715d7302ea82

Subscribe to this mod's changes

internal-knowledge-base CLAUDE.md is an instructions file published in the GitHub repository kimsb2429/internal-knowledge-base (0 stars, last pushed 5mo ago), licensed MIT. It adds 821 tokens to every session, about $0.0033 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

internal-knowledge-base CLAUDE.md next to 6 instructions from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
internal-knowledge-base CLAUDE.mdkimsb2429/internal-knowledge-base	Instructions file	0	5mo ago	MIT	A	821	Claude Code
vscode buildNext.instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	9,719	GitHub Copilot
next.js AGENTS.mdvercel/next.js	Instructions file	143k	yesterday	MIT	A	7,461	Codex, OpenCode
langchain AGENTS.mdlangchain-ai/langchain	Instructions file	149k	yesterday	MIT	A	4,680	Codex, OpenCode
vscode oss-third-party-notices.instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	5,516	GitHub Copilot
vscode copilot-instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	3,093	GitHub Copilot
deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness	Instructions file	274k	2d ago	MIT	A	4,313	Codex, OpenCode

internal-knowledge-base: Instructions file for Claude Code