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/jamestelfer/sandy/claude-mdgit clone --depth 1 https://github.com/jamestelfer/sandyWrote 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/jamestelfer/sandy/claude-md)<a href="https://agentmods.dev/instructions/jamestelfer/sandy/claude-md"><img src="https://agentmods.dev/badge/instructions/jamestelfer/sandy/claude-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 | $0.01475 | $0.01475 |
| Opus 5 | $0.00737 | $0.00737 |
| Sonnet 5 | $0.00295 | $0.00295 |
| Haiku 4.5 | $0.00147 | $0.00147 |
Grade A, and why
sandy 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 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.
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 — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code when working with code in this repository.
Voice
Apply these rules to all prose, commit messages, PR descriptions, and user-facing text.
- Technical-professional register. Use terms directly without over-explanation.
- Use Conventional Commit leads for all commit messages and PR descriptions.
- Guides: second person, imperative mood. Start instructions with the verb.
- Reference: third person, declarative mood. Describe what things are and do.
- No first person ("I", "we") anywhere except
Acknowledgements. - No filler ("Let's dive in", "It's important to note", "As mentioned earlier").
- No hedging ("You might want to consider", "It's generally a good idea").
- No rhetorical questions. State the information instead.
- No apologetic framing ("This might seem complicated, but…").
- Sentences under 20 words where possible. Single-sentence paragraphs are fine.
- British/Australian spelling ("favour", "organise", "colour", "licence" as noun).
What is Sandy
Sandy runs TypeScript scripts inside sandboxed environments (Shuru microVMs or Docker containers) with AWS SDK access via IMDS, for AI agents to safely execute read-only AWS queries without exposing credentials. Built on Bun — uses Bun as runtime, test runner, and binary compiler.
Two entry points: CLI (sandy) for direct use, MCP server (sandy mcp) for AI agent use via Model Context Protocol.
Project Layout
src/core/ Shared types and config persistence
src/sandbox/ Backend interface, implementations, backend factory
src/session/ Session lifecycle and working directory setup
src/resources/ Embedded resources, bootstrap staging, checks, temp dirs
src/execution/ Runtime environment and output scanning utilities
src/output/ Output handling, line writing, progress parsing
src/logging/ Structured logging
src/cli/ CLI parser and command registration
src/cli/commands/ One file per CLI subcommand (config, image, check, run, mcp)
src/mcp/ MCP server with tool/resource registration modules
src/test-support/ Shared test doubles and test helper utilities
embedded/ Files packed into embedded.tar and loaded via memfs at runtime
embedded/bootstrap/ Bootstrap files staged into the sandbox during image creation
embedded/checks/ Baseline and connect check scripts
embedded/skills/ CLI and MCP skill definitions and resources
plans/ Implementation phase plans
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.
- 3d ago First seen · 110 lines · 1,475 tokens per session scan A ff2ecf6d0a23
sandy CLAUDE.md is an instructions file published in the GitHub repository jamestelfer/sandy (7 stars, last pushed 29d ago), licensed Apache-2.0. It adds 1,475 tokens to every session, about $0.0074 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
aws-bench AGENTS.md
AGENTS.md instructions for aws-bench/aws-bench, covering agents.md, setup, commands, project structure and code style.
agentconfig.org AGENTS.md
Instructions for agentconfig/agentconfig.org, covering agent instructions for agentconfig.org, project overview, target audience, site structure and tech stack.
.opencode
Instructions for rshade/finfocus, covering opencode directory instructions, guidelines for this directory, purpose and when working with opencode files.
webview-devtools-mcp AGENTS.md
AGENTS.md instructions for ilharp/webview-devtools-mcp, covering which files you can freely rewrite, the patch stack is managed by hand, repo layout at a glance and build.
meet-ai AGENTS.md
Instructions for SoftWare-A-G/meet-ai, covering instruction for agents to work with meet-ai monorepo, commands, tanstack start — framework & server, tanstack router — routing & navigation and tanstack router — tooling.
piratebao copilot-instructions.md
Instructions for d4551/piratebao: PirateBao kraken active by default. Respond like pirate captain packing dumpling cargo: dense, exact, no wasted wrap.