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/neoswarm/neosh/agents-mdgit clone --depth 1 https://github.com/neoswarm/neoshWrote 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/neoswarm/neosh/agents-md)<a href="https://agentmods.dev/instructions/neoswarm/neosh/agents-md"><img src="https://agentmods.dev/badge/instructions/neoswarm/neosh/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 | $0.18727 | $0.18727 |
| Opus 5 | $0.09363 | $0.09363 |
| Sonnet 5 | $0.03745 | $0.03745 |
| Haiku 4.5 | $0.01873 | $0.01873 |
Grade A, and why
neosh 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 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 — 890 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working on neosh
A terminal-first agent workspace: Rust core, TypeScript plugins, "the extensibility model of Neovim and the feature set of Cursor".
The thesis is that extensibility is the product. Every capability ships on the same public API a third-party plugin author would use. If a feature needs a private escape hatch, the API is wrong and the API is what to fix.
Which means the user owns the workspace, not us: every part of it is theirs to replace, and a plugin of their own is how they do it. New windows, panels and floats; new commands and keys on any of them; different colours, different layout, a sidebar that is nothing like ours — none of that is a feature we grant, it is what the API is for. Ours are the defaults, and a default is something you turn off.
Layout
| Crate | Owns |
|---|---|
neosh-proto |
Every wire type. #[derive(TS)] on all of them; generated .ts is committed. |
neosh-core |
Buffers, windows, floats, extmarks, highlights, focus, keymaps, the single-writer Editor. No I/O. |
neosh-script |
The deno_core host and the op surface. The only crate that knows deno exists. |
neosh-provider |
trait Provider, the shared SSE parser, and one driver per transport. |
neosh-agent |
Session, Turn, the tool registry, the permission layer, the agent loop. |
neosh-syntax |
Syntax highlighting: what a token is, never what colour it is. The only crate that knows a grammar exists. |
neosh-tui |
Terminal frontend. Owns all display-width maths. |
neosh |
The binary. Wires the tasks, owns config, hosts the chat — and is both the workspace (daemon.rs) and the terminal that views it (client.rs). |
neosh-plugins |
plugins/ itself. One file of Rust holding the include_dir!s, because the tree an include_dir! reads has to be under the package root or it embeds here and publishes empty. |
plugins/api/ is @neosh/api — generated protocol types plus a hand-written ergonomic layer.
plugins/builtin/ are ordinary plugins that happen to ship in the binary. The directory is also a
crate (plugins/Cargo.toml, plugins/lib.rs) and nothing moved to make it one: cargo package
carries only what is under a package root, so a neosh-script embedding ../../plugins built fine
in a checkout and would have uploaded a crate with no plugins and no API source in it. Publishing
is docs/releasing.md.
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 · 890 lines · 18,727 tokens per session scan A 13d036b77722
neosh AGENTS.md is an instructions file published in the GitHub repository neoswarm/neosh (22 stars, last pushed 4d ago), licensed MIT. It adds 18,727 tokens to every session, about $0.0936 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
hindsight CLAUDE.md
Instructions for vectorize-io/hindsight, covering claude.md, project overview, development commands, local development (api + ui) and start both api server and control plane ui.
codedb AGENTS.md
AGENTS.md instructions for justrach/codedb, covering codedb agent guidelines, what codedb is (and isn't), review guidelines, pre-merge verification and security-sensitive areas.
claude-code-settings copilot-instructions.md
Instructions for feiskyer/claude-code-settings, covering claude.md, environment setup, required dependencies, configuration and skills.
valuecell AGENTS.md
Instructions for ValueCell-ai/valuecell, covering guidelines, python programming, python environment, imports and runtime checks.
neeko AGENTS.md
Instructions for tincopper/neeko, covering neeko — repository guidelines, 项目概览, 目录结构, 前端（feature-based 架构） and 后端（domain-driven 模块化架构）.
pi CLAUDE.md
Instructions for share-skills/pi, covering 强制, 内部目标（不可写入公开文件）, 跑分规范 and eval 驱动优化规则.