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/kasssandr/archilles/agents-mdgit clone --depth 1 https://github.com/kasssandr/archillesWrote 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.
[](https://agentmods.dev/instructions/kasssandr/archilles/agents-md)<a href="https://agentmods.dev/instructions/kasssandr/archilles/agents-md"><img src="https://agentmods.dev/badge/instructions/kasssandr/archilles/agents-md.svg" alt="Measured on agentmods" height="20"></a>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 | $0.01256 | $0.01256 |
| Opus 5 | $0.00628 | $0.00628 |
| Sonnet 5 | $0.00251 | $0.00251 |
| Haiku 4.5 | $0.00126 | $0.00126 |
Grade A, and why
archilles AGENTS.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 5d 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.
How it starts
The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file provides guidance to Codex when working with code in this repository.
Project Overview
ARCHILLES is a privacy-first, local-first RAG (Retrieval-Augmented Generation) system for semantic search across Calibre e-book libraries. It provides hybrid vector+BM25 search with academic-grade citations, MCP integration for Claude Desktop; it will provide MCP integration for other LLMs soon, and finally a Streamlit web UI.
Note for Codex: ARCHILLES cannot currently be consumed via MCP from Codex — local MCP servers are not supported there. If that policy changes, wiring it up is worth a new attempt.
Commands
Installation
pip install -r requirements.txt
Indexing
python scripts/rag_demo.py index "/path/to/book.pdf" --book-id "AuthorName"
python scripts/batch_index.py --tag "Your-Tag" [--dry-run] [--skip-existing]
Searching
python scripts/rag_demo.py query "search term" [--mode hybrid|semantic|keyword]
python scripts/rag_demo.py query "text" --language de --tag-filter History
python scripts/rag_demo.py stats
python scripts/rag_demo.py list-indexed
Web UI
python scripts/web_ui.py
MCP Server (Claude Desktop)
python mcp_server.py
Code Quality
black src/ scripts/ # formatting
flake8 src/ scripts/ # linting
pytest # tests
Architecture
The system follows a layered pipeline:
Calibre Library (read-only SQLite)
↓
Extractors (PDF/EPUB/TXT/HTML/MOBI/DJVU/OCR)
↓
Modular Pipeline: ParserRegistry → ChunkerRegistry → EmbedderRegistry
↓
LanceDB (hybrid: dense vectors + BM25 FTS)
↓
Retriever (RRF fusion + optional cross-encoder reranking)
↓
ArchillesService (central facade)
↓
Consumers: MCP Server | Web UI (Streamlit) | CLI
Key Modules
src/calibre_db.py— Read-only access to Calibre'smetadata.db(SQLite). This is an absolute boundary: never write to the Calibre library.src/extractors/— Format-specific extractors. PDF uses PyMuPDF primary with pdfplumber fallback. EPUB uses ebooklib with TOC-based section classification.src/archilles/pipeline.py—ModularPipelineorchestrates Parser → Chunker → Embedder using registry-based components.src/storage/lancedb_store.py— LanceDB backend. Stores 1024-dim BGE-M3 vectors with rich metadata. Two tables:chunks(main content) andannotations(user highlights/notes).src/service/archilles_service.py— Single facade used by MCP server, web UI, and CLI. Start here when adding new features.src/calibre_mcp/server.py— MCP server exposing 10 tools (search, metadata, citations, annotations, stats). Carefully manages stdout/stderr to avoid JSON-RPC protocol corruption.mcp_server.py— Entry point for Claude Desktop MCP integration.scripts/rag_demo.py— Primary CLI (2,479 lines). Index, query, stats, list-indexed.
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.
- 5d ago First seen · 119 lines · 1,256 tokens per session scan A 8804deaa9933
archilles AGENTS.md is an instructions file published in the GitHub repository kasssandr/archilles (8 stars, last pushed today), licensed MIT. It adds 1,256 tokens to every session, about $0.0063 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
codesearch AGENTS.md
AGENTS.md instructions for flupkede/codesearch, covering agents.md — codesearch, current state, implemented features, ⚠️ design constraint — never poll a federated peer and open todos.
localmind AGENTS.md
Instructions for shaul1991/localmind, covering agents.md — localmind 작업 규약, 최상위 전제 — 왜 이 프로젝트가 존재하는가 (2026-07-23), 범위 우선순위 — 코어 우선·메타 동결 (2026-07-21), 작업 방법론 — 애드온이 정본 (2026-08-07) and 원격 반영 — 브랜치·pr·ci.
localmind GEMINI.md
Instructions for shaul1991/localmind: 정본은 AGENTS.md — Claude Code·Codex·Gemini CLI가 같은 규약을 따르도록 한 곳에 유지한다.
localmind CLAUDE.md
Instructions for shaul1991/localmind: 정본은 AGENTS.md — Claude Code와 Codex가 같은 규약을 따르도록 한 곳에 유지한다.
codesearch CLAUDE.md
Claude Code instructions for flupkede/codesearch, a project described as: Multi-repo semantic code search MCP server in Rust — hybrid vector + BM25 retrieval, tree-sitter AST chunking, fully offline. For OpenCode, Claude Code, Cursor, and any MCP client.
devrag CLAUDE.md
Claude Code instructions for tomohiro-owada/devrag, covering devrag - claude code project guide, quick reference, project structure, key components and 1. configuration (internal/config/).