engraphis CLAUDE.md

Repository-specific instructions for Claude Code, a coding agent, covering project structure, working rules, available commands, and required checks. It directs the agent to use the repository's AGENTS.md file as the main manual.

In plain words
What is it for?
Working in the Engraphis repository, deciding whether changes belong to the newer or legacy codebase, running the canonical offline test gate, and updating evaluations for retrieval or ranking changes.
Why use it?
It helps the agent choose the correct codebase, follow the project's conventions, and run the required checks before declaring work complete.

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/coding-dev-tools/engraphis/claude-md
Clone the repo
git clone --depth 1 https://github.com/Coding-Dev-Tools/engraphis
Per session 1,082 This file is loaded in full into every session.
When invoked 1,082 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 $0.01082 $0.01082
Opus 5 $0.00541 $0.00541
Sonnet 5 $0.00216 $0.00216
Haiku 4.5 $0.00108 $0.00108

Measured yesterday against content hash 87de494e009e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

engraphis 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.

CLAUDE.md · 72 lines

How it starts

The opening of the file, as written. The whole thing — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

The full operating manual for this repo lives in @AGENTS.md — read it first. It is the canonical, vendor-neutral source (architecture, commands, conventions, algorithms, gotchas); everything there applies to Claude Code. This file adds only Claude-specific guidance, so the two never drift.

The one rule that prevents most mistakes

Two codebases share engraphis/: v2 (core/ + backends/ — the target) and the v1 legacy FastAPI server (app.py, routes/, stores/, engines/, flat namespaces). Build new capability on v2 behind the interfaces in core/interfaces.py. Decide which side a change belongs to before editing. Full table: AGENTS.md §0.

Before you say "done" — run the canonical gate

Use the exact primary offline gate in AGENTS.md §1; do not maintain a smaller duplicate here. .github/workflows/ci.yml is authoritative for the current Python matrix and dedicated typecheck, encryption, and built-artifact jobs. No network or API key is required for the primary gate. If you changed retrieval, scoring, or ranking, add or update an eval—per AGENTS.md §3.7, "better" needs a number, not an assertion.

Slash commands available here

  • /init — regenerate codebase documentation.
  • /review — review a GitHub pull request (/code-review for the local working diff).
  • /security-review — review pending changes for vulnerabilities. Run this before finishing any change to the write/ingest path: ingested content is treated as untrusted and memory poisoning is an explicit threat. This now also covers MemoryEngine.index_repo() (reads local files at an agent-supplied path — see SECURITY.md §5) and the deterministic conflict resolver (core/resolve.py).

Working style in this repo

  • Interface-first & dependency-light (AGENTS.md §3): every core/ module, including core/engine.py, remains protocol-only and runnable on NumPy. Concrete backend selection lives in the outer composition root engraphis/factory.py; engraphis/__init__.py registers it for MemoryEngine.create(), and engraphis.create_memory_engine() exposes it directly. Gate heavy imports behind backend factories and never import a concrete backend from core/.
  • House style: ruff line-length 100, Python 3.9-compatible syntax, pure/tested scoring functions, provenance and scope on every memory.
  • Be concise and direct in chat — explain the why of a change briefly, link the file, and let the diff speak.
  • When code and docs disagree, the code wins — then fix the doc in the same change, and update AGENTS.md/CLAUDE.md if a convention or command changed.

Read the full file on GitHub · 72 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. yesterday First seen · 72 lines · 1,082 tokens per session scan A 87de494e009e

Subscribe to this mod's changes

engraphis CLAUDE.md is an instructions file published in the GitHub repository Coding-Dev-Tools/engraphis (170 stars, last pushed 2d ago), licensed Apache-2.0. It adds 1,082 tokens to every session, about $0.0054 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 2026-08-30.