llm-wiki AGENTS.md

Instructions for an AI agent that turns source documents into a connected wiki, a collection of linked Markdown articles that people can search and read.

In plain words
What is it for?
Use it to ingest source material, build and update a knowledge base, and query the resulting wiki from its configured location.
Why use it?
It gives the agent a defined process for organizing scattered information without requiring a person to maintain every article by hand.

Instructions file for CodexOpenCode

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/nvk/llm-wiki/agents-md
Clone the repo
git clone --depth 1 https://github.com/nvk/llm-wiki

Made for: Codex, OpenCode.

Per session 13,715 This file is loaded in full into every session.
When invoked 13,715 The same file — it is already loaded in full.
Security scan A 1 finding. 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 $0.13715 $0.13715
Opus 5 $0.06857 $0.06857
Sonnet 5 $0.02743 $0.02743
Haiku 4.5 $0.01371 $0.01371

Measured 3d ago against content hash 4ce9eaf3607a, 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 AGENTS.md scanned grade A with 1 finding 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

and IPv4 retry (`curl -4`) when media hosts hang on IPv6.
AGENTS.md · 1,060 lines

How it starts

The opening of the file, as written. The whole thing — 1,060 lines — stays where its author put it; the contents beside it link to each section on GitHub.

LLM Wiki — Agent Instructions

This is an "idea file" in the spirit of Karpathy's LLM Wiki. Paste it into any LLM agent (OpenAI Codex, Claude Code, OpenCode, Gemini Code Assist, or similar) and it will build and manage a wiki for you. The agent customizes the specifics; this file communicates the system.

Editing the llm-wiki repo itself? This file is the portable wiki protocol for end users. The dev contract for the plugin codebase (testing, sync workflow, project structure, symlink invariant) lives in CLAUDE.md — read it in addition to this file. Both Claude Code and Codex agents working on the repo should treat CLAUDE.md as the source of truth for repo-level workflow.

What This Is

An LLM-compiled knowledge base. You (the LLM agent) are both the compiler and the query engine. Raw source documents are ingested, then incrementally compiled into a wiki of interconnected markdown articles. The human rarely edits the wiki directly — that's your domain.

The metaphor: raw sources are source code, you are the compiler, the wiki is the executable.

Architecture

Hub Path

Defaults to ~/wiki/. Optionally configured via ~/.config/llm-wiki/config.json:

{ "hub_path": "~/Library/Mobile Documents/com~apple~CloudDocs/wiki" }

If the config exists, prefer hub_path (expand only the leading ~) instead of ~/wiki/ everywhere below. Treat resolved_path from older configs as a machine-specific fallback cache, not as the source of truth; do not write it into shared configs. If the path can be statted but reading wikis.json or listing topics/ fails with Operation not permitted, the path is correct and macOS/iCloud privacy is blocking the launcher; ask the user to grant Full Disk Access or iCloud Drive access and restart, rather than falling back to another machine path. Store the selected path as HUB. Use /wiki config hub-path <path> to set it up, or create the JSON manually.

Read the full file on GitHub · 1,060 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 · 1,060 lines · 13,715 tokens per session scan A 4ce9eaf3607a

Subscribe to this mod's changes

llm-wiki AGENTS.md is an instructions file published in the GitHub repository nvk/llm-wiki (1,175 stars, last pushed 6d ago), licensed MIT. It adds 13,715 tokens to every session, about $0.0686 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.