What is limbic CLAUDE.md?

Project instructions for Limbic, a data-curation toolkit for embeddings, search, change proposals, and verification. Embeddings are numerical representations of text used to compare meaning.

Use them when working on Limbic data processing, especially whitening focused corpora, generalizing changing numbers, and choosing clustering thresholds.

They provide rules intended to make similarity search and clustering more reliable for focused or number-heavy text collections.

How houshuang/limbic configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to houshuang/limbic. 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/houshuang/limbic/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/houshuang/limbic

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/houshuang/limbic/claude-md/github.svg)](https://agentmods.dev/instructions/houshuang/limbic/claude-md)

Your own site

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

Per session 3,255 This file is loaded in full into every session.

When invoked 3,255 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	$0.03255	$0.03255
Opus 5.5	$0.01302	$0.01302
Sonnet 5.5	$0.00651	$0.00651
Haiku 4.5	$0.00326	$0.00326

Measured 1mo ago against content hash 8cd6050d61c7, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

limbic 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 1mo 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 · 252 lines

How it starts

The opening of the file, as written. The whole thing — 252 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Limbic — AI Agent Guide

Data curation toolkit: embeddings, search, proposals, and AI-assisted verification.

Three core packages: limbic.amygdala (find patterns), limbic.hippocampus (manage changes), limbic.cerebellum (verify correctness) — plus limbic.drive, a planning policy that runs before any of them.

Rules of Thumb

Always whiten domain-focused corpora

If the corpus is about one domain (education, medicine, politics, performing arts), always use whitening. Without it, raw embeddings compress into a narrow similarity band (0.7–0.9) and downstream clustering/novelty/search all degrade.

model = EmbeddingModel(whiten_epsilon=0.1)
model.fit_whitening(corpus_texts)

Skip whitening only when your corpus spans many unrelated domains.

Always genericize number-heavy text

If texts contain variable amounts, dates, section references, or place names around the same argument, use genericize=True. This prevents "allocate 50M" and "allocate 200M" from being treated as different arguments.

model = EmbeddingModel(genericize=True, whiten_epsilon=0.1)

Clustering thresholds depend on whitening

Corpus state	Threshold	Why
Raw embeddings, diverse corpus	0.70–0.75	Embeddings already spread
Raw embeddings, domain-focused	0.90+	Narrow cone, everything looks similar
Whitened, homogeneous text (extracted claims)	0.85	Very similar surface form
Whitened, diverse authorship (op-eds, responses)	0.70–0.75	Different writers phrase same argument differently

If your largest cluster has 50+ members, your threshold is too low or you need whitening. If you get zero clusters, it is too high — whitening can leave a corpus whose maximum pairwise similarity is below 0.85. Start at 0.75 post-whitening, then sweep [0.70, 0.75, 0.80, 0.85] on your data. Validated on 27K education claims (0.85), 6.5K political proposals (0.75–0.80), and 1.7K op-ed claims (0.75).

Always validate thresholds before shipping

Read the full file on GitHub · 252 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.

1mo ago Changed · +53 lines · +815 tokens per session 8cd6050d61c7
1mo ago Changed · +29 lines · +507 tokens per session 131540a8c807
2mo ago First seen · 170 lines · 1,933 tokens per session scan A 1d57fd6ffad7

Subscribe to this mod's changes

limbic CLAUDE.md is an instructions file published in the GitHub repository houshuang/limbic (3 stars, last pushed 4d ago), licensed MIT. It adds 3,255 tokens to every session, about $0.0130 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.

Clustering Data and AI embeddings knowledge-management nlp novelty-detection personal-knowledge

limbic CLAUDE.md next to 6 instructions from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
limbic CLAUDE.mdhoushuang/limbic	Instructions file	3	4d ago	MIT	A	3,255	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

limbic: Instructions file for Claude Code