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/cq27-dev/rag-rat/agents-mdgit clone --depth 1 https://github.com/cq27-dev/rag-ratWhat 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 | $0.02971 | $0.02971 |
| Opus 5 | $0.01486 | $0.01486 |
| Sonnet 5 | $0.00594 | $0.00594 |
| Haiku 4.5 | $0.00297 | $0.00297 |
Grade A, and why
rag-rat 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 2d 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 — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Guidance for coding agents working in the rag-rat repository.
(CLAUDE.md is a symlink to this file.)
Prefer the rag-rat MCP for code browsing
This repo ships rag-rat — a local repo-intelligence index and MCP server — and it is indexed by
its own server (dogfooded). Heavily prefer the rag-rat MCP tools over raw grep/cat/file
reads when browsing or understanding code. One MCP call returns more context, faster, than a
shell sweep, and it surfaces drive-by repo memories (source-anchored invariants, decisions,
risks) attached to the code you're touching — context you would otherwise never see.
Reach for these first:
semantic_search— "where is this concept implemented?" Returns current source chunks with inline graph (callers/callees), git, and GitHub papertrail, all validated against current source.symbol_lookup— exact/fuzzy symbol resolution (Rust/TS/Kotlin/C/C++/Python/Swift/Go), with any bound memories attached.impact_surface— the coding preflight before editing a symbol: graph callers/callees, tests, git history, papertrail, and repo memories crossing the call path. Run it before changing anything non-trivial.find_callers/trace_callees— reverse/forward graph traversal instead of grepping for call sites.read_chunk— current text for a chunk with anchor validation + graph + memories.repo_brief/repo_clusters— orientation (spine, churn, god-modules, ownership clusters).important_symbols— load-bearing symbols by (SCIP-aware) PageRank; passpersonalizeto bias toward what you're editing. Compiler-grade oncerag-rat oracle runhas run.
Symbol handle: symbol-returning tools emit id, an opaque sym_<hex> token — the stable handle
to cache and pass back into graph/impact/memory tools as the id param (copy verbatim; never parse
it as a number). There is no numeric symbol_id on the wire (it's an internal rowid reassigned on
every reindex). Use ref (the path::name qualified name) for the human-readable identity. The
symbol-tool params are ref / id / lang (formerly symbol_path / logical_symbol_id /
language).
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.
- 2d ago First seen · 185 lines · 2,971 tokens per session scan A b728f0f35048
rag-rat AGENTS.md is an instructions file published in the GitHub repository cq27-dev/rag-rat (17 stars, last pushed 4d ago), licensed MIT. It adds 2,971 tokens to every session, about $0.0149 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.
Other instructions, from other repositories
jcode-ragmir AGENTS.md
AGENTS.md instructions for jcode-works/jcode-ragmir, covering ragmir, working rules, product boundary, privacy and ingestion and documentation.
rag-vault AGENTS.md
AGENTS.md instructions for maxrotemberg04-spec/rag-vault, covering rag-vault — agent instructions and hard rules.
grounding-ai CLAUDE.md
Claude Code instructions for andyliszewski/grounding-ai, covering claude.md, critical requirements, python version, embedding generation is mandatory and after ingestion, update embeddings for affected agents.
siyuan AGENTS.md
AGENTS.md instructions for siyuan-note/siyuan, covering agents.md, 1. non-negotiable constraints, do not hand-edit, verification and prohibited operations and 2. project-specific rules.
veritas-kanban AGENTS.md
Instructions for BradGroux/veritas-kanban, covering agents.md — canonical agent instructions for veritas kanban, runtime requirements, repository layout, essential commands and install.
GPT-RAG release.instructions.md
Instructions for Azure/GPT-RAG, a project described as: Sharing the learning along the way we been gathering to enable Azure OpenAI at enterprise scale in a secure manner. GPT-RAG core is a Retrieval-Augmented Generation pattern running in Azure, using Azure Cognitive Search for retrieval and Azure OpenAI large…