llm-wiki CLAUDE.md

llm-wiki CLAUDE.md is an instructions file for coding agents from krishddd/llm-wiki. It costs 7,304 tokens per session, scanned A, original, MIT.

A set of instructions for llm-wiki, a wiki that stores documents, model settings, searches, citations, and durable research notes.

In plain words
What is it for?
It helps configure model providers, search the wiki, ingest files, save researched answers, manage embeddings, and run maintenance jobs.
Why use it?
It defines which language models handle different jobs and how information is retrieved, checked, saved, and periodically cleaned up.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/krishddd/llm-wiki/claude-md
Clone the repo
git clone --depth 1 https://github.com/krishddd/llm-wiki

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.

agentmods badge for llm-wiki CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/krishddd/llm-wiki/claude-md.svg)](https://agentmods.dev/instructions/krishddd/llm-wiki/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/krishddd/llm-wiki/claude-md"><img src="https://agentmods.dev/badge/instructions/krishddd/llm-wiki/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 7,304 This file is loaded in full into every session.
When invoked 7,304 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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.

ModelPer sessionOnce invoked
Fable 5 $0.07304 $0.07304
Opus 5 $0.03652 $0.03652
Sonnet 5 $0.01461 $0.01461
Haiku 4.5 $0.00730 $0.00730

Measured 3d ago against content hash e9238632573e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

llm-wiki 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 3d 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 · 472 lines

How it starts

The opening of the file, as written. The whole thing — 472 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — LLM-Wiki v2 Schema

Read by Claude Code at the start of every session. Defines the wiki architecture, memory tiers, conventions, and workflows. Update this file when conventions change. Companion AGENTS.md is the agent-facing tool catalogue.


Models

Role Model Strengths
Summarise, extract gemma4:e4b Fast, strong instruction-following
Reason, route, lint, claims qwen3:14b Deep reasoning, thinking mode
Quantitative reasoning specialist vibethinker:3b AIME-class maths / STEM / code; routed to ADAPTIVELY for quantitative questions
Embeddings nomic-embed-text:latest 274 MB, MTEB-strong
STEM embeddings (optional) bge-m3 Domain-routed embedder for maths/science/econ/eng; enable with EMBED_STEM_ENABLED
Vision (image captions) llava:7b Optional — used when ingest_caption_images=true

All models served via Ollama at OLLAMA_HOST (default http://localhost:11434). MODEL_FAST = MODEL_REASON is intentional — disables a missing-llama3.2 fallback.

Multi-provider LLM fleet (v4, llm_wiki/providers.py)

Each role can be routed to a hosted, OpenAI-compatible provider instead of Ollama via PROVIDER_<ROLE> (summary / reason / fast / solver / embed / vision), default ollama. Vision routes the llava image-caption role to a multimodal provider via the OpenAI image_url content schema. Providers (v6 roster, PROVIDER_DEFS registry): groq, github (GitHub Models), gemini, openai, anthropic (Claude via api.anthropic.com/v1 OpenAI-compat; force_max_tokens), xai (Grok), openrouter (100+ OSS models, one key), and custom — any OpenAI-compatible gateway (vLLM / LM Studio / llama.cpp / Together / DeepSeek …) via CUSTOM_BASE_URL/CUSTOM_MODEL (key optional for local gateways). Embeddings: ollama | gemini | openai | custom only. resolve_chat_provider() returns None (→ Ollama) when the role is ollama or its key/model is unset, so routing is opt-in and self-healing. A provider HTTP error is re-raised as OllamaError so the existing role fallbacks (e.g. reason→fast) still fire. Keys are read from env ONLY — never committed. .env.example documents every knob.

Read the full file on GitHub · 472 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.

  1. 3d ago First seen · 472 lines · 7,304 tokens per session scan A e9238632573e

Subscribe to this mod's changes

llm-wiki CLAUDE.md is an instructions file published in the GitHub repository krishddd/llm-wiki (2 stars, last pushed 11d ago), licensed MIT. It adds 7,304 tokens to every session, about $0.0365 per session on Opus 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.