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/sandbaseai/sandbase-harness/agents-mdgit clone --depth 1 https://github.com/sandbaseai/sandbase-harnessWrote 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/sandbaseai/sandbase-harness/agents-md)<a href="https://agentmods.dev/instructions/sandbaseai/sandbase-harness/agents-md"><img src="https://agentmods.dev/badge/instructions/sandbaseai/sandbase-harness/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 | $0.01731 | $0.01731 |
| Opus 5 | $0.00865 | $0.00865 |
| Sonnet 5 | $0.00346 | $0.00346 |
| Haiku 4.5 | $0.00173 | $0.00173 |
Grade A, and why
sandbase-harness 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 4d 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 — 187 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SandBase Harness Agent Instructions
Project identity
SandBase Harness (managed-agents) is a local-first, self-hosted runtime for
AI agents. It provides a Claude Managed Agents-style /v1 API, a local
Console, persistent sessions, resumable event streams, SQLite state, memory,
skills, credential vaults, MCP toolsets, audit/replay, snapshots, a TypeScript
SDK, and local, Docker, Kubernetes, and self-hosted sandbox providers.
This is an independent SandBase open-source project. It is not an official Anthropic, DeepSeek, or DSH implementation. The DeepSeek Harness Handbook is a related community documentation project and may be linked when relevant, but the two products must be described separately.
Repository map
src/api/: HTTP routes and protocol adapters.src/core/session/: lifecycle, events, recovery, context, tools, snapshots.src/strategy/: model-loop implementations and stream handling.src/sandbox/: execution providers and capability declarations.src/core/db/: SQLite connection and embedded migrations.src/core/runtime/: bootstrap and service composition.apps/console/: React/Vite operator console.src/sdk/: TypeScript client SDK.tests/unit/,tests/integration/: regression and behavior coverage.docs/,examples/: specifications, guides, and examples.
The main execution path is:
API/SDK → SessionManager → ContextBuilder → AgentStrategy → Model/MCP tools → Sandbox
The local sandbox is not a security boundary. Untrusted agent code must use an isolated provider such as Docker or Kubernetes.
Role A: project maintenance
Act as the project owner and maintainer. Within the repository scope, handle routine safe work autonomously.
Issue workflow
- Inspect
git status --short, recent commits, the README, repository layout, open Issues, open PRs, and recent releases before choosing work. - Triage each Issue using source evidence. Reproduce when possible, identify the failing boundary, detect duplicates, and leave a concise factual comment.
- Create focused branches named
fix/issue-<number>-<slug>for bugs orfeat/issue-<number>-<slug>for features. - Implement the smallest complete fix and add regression tests. Update docs, migrations, and changelog entries when public behavior changes.
- Link the PR with
Fixes #<number>orCloses #<number>only when it genuinely resolves that Issue.
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.
- 4d ago First seen · 187 lines · 1,731 tokens per session scan A 1406388d8562
sandbase-harness AGENTS.md is an instructions file published in the GitHub repository sandbaseai/sandbase-harness (642 stars, last pushed 3d ago), licensed Apache-2.0. It adds 1,731 tokens to every session, about $0.0087 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
OpenQuantum AGENTS.md
AGENTS.md instructions for xi-zhao/OpenQuantum, covering openquantum, 四个职责面, 扩展对象不可混用, 不变量 and 代码位置.
melra CLAUDE.md
Instructions for XAGI-Lab/melra, covering claude.md, what this is, commands, execution pipeline (the core invariant) and making changes.
openrouter-mcp-multimodal AGENTS.md
AGENTS.md instructions for stabgan/openrouter-mcp-multimodal, covering agent instructions, before you ship, releasing (read this before publishing), short version and version files (must all match package.json).
Plonk AGENTS.md
AGENTS.md instructions for ostapondo/Plonk, covering agent rules, layout, adding a module, build & verify and code style.
mcp-dockhand copilot-review.instructions.md
Copilot instructions for strausmann/mcp-dockhand: When reviewing pull requests in this repository.
minecraft-dev-mcp AGENTS.md
Instructions for MCDxAI/minecraft-dev-mcp, covering agent handbook – minecraft dev mcp, project snapshot, what agents should prioritize, architecture wayfinder (src/) and available mcp tools (for llm surfaces).