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.
npx agentmods add instructions/lordymine/codegraph/claude-mdgit clone --depth 1 https://github.com/Lordymine/codegraphWhat 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 | $0.02284 | $0.02284 |
| Opus 5 | $0.01142 | $0.01142 |
| Sonnet 5 | $0.00457 | $0.00457 |
| Haiku 4.5 | $0.00228 | $0.00228 |
Grade A, and why
codegraph 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 yesterday.
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.
How it starts
The opening of the file, as written. The whole thing — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — codegraph
Trabalhe SEMPRE a partir desta pasta (
d:/projetos/codegraph), abrindo a sessão do Claude Code aqui dentro — não pela pasta do ajudaqui. Este projeto é independente, com git próprio.
O que é
codegraph — grafo de conhecimento de código token-efficient para agentes de
IA, em Go. Reimplementação enxuta das ideias do
DeusData/codebase-memory-mcp
(que é C puro). Foco no stack do Rafael (TypeScript/JS + Go + NestJS), não 158
linguagens. Module: github.com/Lordymine/codegraph. Go 1.26.
Indexa um repo → grafo SQLite de símbolos + relações; o agente consulta o grafo (quem-chama, callers/callees, busca rankeada) em vez de ler arquivo por arquivo.
Princípios de design (não violar sem motivo)
- Storage é trivial — 2 tabelas (
nodes,edges) + FTS5, igual ao upstream. O valor está nas arestas CALLS, não no banco. - Token-efficient por construção — toda query retorna ref compacta
(
name + qualified_name + label + file + line), NUNCA código. Código só via a toolsnippet. É daí que vem o ~10× de economia de token. - Aposta-chave: delegar ao type-checker da linguagem, não reimplementar. O
upstream refez resolução de tipos em C ("Hybrid LSP", ~9 langs, meses de
trabalho). Nós delegamos a resolução de chamadas aos indexadores batch da
linguagem —
scip-typescript(TS/JS) ego/packages+go/callgraph(Go, as libs que o próprio gopls usa) — lidos in-process. Mesma precisão de type-checker real, num passe só, sem servidor LSP vivo pra babá. Pulamos a parte mais difícil do port. - Precisão honesta — aresta não resolvida é descartada (endpoints precisam existir no grafo). Melhor aresta faltando que aresta errada.
- Poucas dependências — SQLite puro-Go (
modernc.org/sqlite), MCP server em stdlib. tree-sitter (cgo) é a dep pesada do M1 — por isso o build exigeCGO_ENABLED=1+ gcc. scip-typescript entra como ferramenta de build (Node), não liga no binário.
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.
- yesterday First seen · 133 lines · 2,284 tokens per session scan A ebdc2062108e
codegraph CLAUDE.md is an instructions file published in the GitHub repository Lordymine/codegraph (5 stars, last pushed 2mo ago), licensed MIT. It adds 2,284 tokens to every session, about $0.0114 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.
Other instructions, from other repositories
kivgraph AGENTS.md
Instructions for Luqueee/kivgraph, covering instrucciones de desarrollo de kivgraph, mapa de instrucciones, identidad del proyecto, qué pregunta contesta cada tool de kivgraph and la puerta delante de grep.
kivgraph CLAUDE.md
Instructions for Luqueee/kivgraph, a project described as: A local MCP server for cross-repository semantic code intelligence in TypeScript and Go, backed by a persistent LadybugDB graph.
CodeStory AGENTS.md
Instructions for TheGreenCedar/CodeStory, covering codestory agent guide, start here, ownership boundaries, product invariants and identity and configuration.
CodeStory copilot-instructions.md
Instructions for TheGreenCedar/CodeStory: Use CodeStory proactively for repository questions. Do not wait for the user to mention it by name.
kiwifs AGENTS.md
AGENTS.md instructions for kiwifs/kiwifs, covering agents.md, project overview, repository structure, build & run and full build (ui + go binary).
lsp-tools copilot-instructions.md
Instructions for zircote/lsp-tools, covering copilot instructions, priorities, key files, commands and when adding language support.