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/grasshopper-protocol/grsshppr/agents-mdgit clone --depth 1 https://github.com/grasshopper-protocol/grsshpprWhat 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 | $0.02083 | $0.02083 |
| Opus 5 | $0.01042 | $0.01042 |
| Sonnet 5 | $0.00417 | $0.00417 |
| Haiku 4.5 | $0.00208 | $0.00208 |
Grade A, and why
grsshppr 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 — 169 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — How Humans and Agents Operate in Grasshopper
Grasshopper is a fully open product: open code, open specs, open design, open decisions, public roadmap. This file is the contract for everyone who changes the project — human or AI. It is enforceable. If a change violates it, the change is wrong, not the rule.
For engineering specifics (tech stack, data model, naming, module rules) see ENGINEERING.md. For visual direction see DESIGN.md and design/principles.md. This file governs how change happens.
1. Principles (non-negotiable)
- Open by default. Publish the reasoning, not just the result. If there is no concrete reason to keep something private, it is public.
- Document before building. Non-trivial change starts as a written proposal (RFC), not as a pull request.
- Decisions must be traceable. Every decision links idea → discussion → decision (ADR) → implementation (PR) → release. No orphan changes.
- Product, Design, and Engineering are equals. A design or product proposal is a first-class contribution, not a request for engineering.
- Lazy stays lazy. This process governs change, not effort. A typo or a one-line fix needs no ceremony. The bar scales with blast radius.
2. Responsibilities of Agents
When acting on this repository, an agent MUST:
- Write an RFC before any major change. "Major" = new feature/module, schema or data-model change, new dependency, public API change, design-system change, governance change, or reversing a prior ADR.
- Link every change to an issue or spec. A PR with no linked issue/RFC is incomplete.
- Record decisions as ADRs. When a choice is made between alternatives, write the ADR (context, options, decision, consequences) and link it.
- Update documentation in the same change. Code that changes behavior updates the relevant doc (README, ENGINEERING, DESIGN, roadmap) in the same PR.
- Ship a test with non-trivial logic. A PR that adds or changes non-trivial
logic includes a test in the same change and must not lower the coverage gate
(see ADR-0011 and
ENGINEERING.md). Logic that isn't unit-testable in place is
extracted into a pure helper under
src/liband tested there. - Verify green before pushing. Run
pnpm verify(typecheck + lint) before pushing a branch. A branch that failspnpm typecheckmust not be pushed — typecheck is exactly whatnext build(Vercel) enforces. The.githooks/pre-commithook runs it automatically as a hard gate; lint stays advisory until the current lint debt is cleared. See ADR-0012. - Resolve merge conflicts cleanly. After resolving any conflict, re-run
pnpm verifyand scan the diff for accidentally duplicated blocks before committing. Never commit both sides of a hunk — a duplicated definition breaks the build. - Record any gate bypass.
git commit --no-verifyis for emergencies only and must be called out in the PR description. A silent bypass is a silent change (see §4). - Never introduce silent changes. No undocumented features, no quiet dependency additions, no schema drift, no behavior changes "while I was in there." If it wasn't asked for and isn't documented, don't ship it.
- Surface operational risk in the repo, not in private notes. Incidents,
drift, and security concerns belong in
/decisionsor SECURITY.md, where humans can see them.
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 · 169 lines · 2,083 tokens per session scan A eeb59003c695
grsshppr AGENTS.md is an instructions file published in the GitHub repository grasshopper-protocol/grsshppr (2 stars, last pushed 2d ago), licensed MIT. It adds 2,083 tokens to every session, about $0.0104 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
postiz-app CLAUDE.md
Claude Code instructions for gitroomhq/postiz-app: This project is Postiz, a tool to schedule social media and chat posts to 28+ channels. You can add posts to the calendar, they will be added into a workflow and posted at the right time. You can find things like.
postiz-app copilot-instructions.md
Copilot instructions for gitroomhq/postiz-app, covering copilot coding agent instructions for postiz, project architecture, developer workflows, conventions & patterns and integration points.
FounderOS-DEMO CLAUDE.md
Instructions for Bennettxai/FounderOS-DEMO, covering founder os, commands, stack, architecture: larp-first, real-ready and g-brain — answered (2026-06-11).
velobase-harness AGENTS.md
Instructions for velobase/velobase-harness, covering ai agent guide, task router, integration router, core rules and general.
kan AGENTS.md
AGENTS.md instructions for kanbn/kan, covering agents.md, project overview, setup commands, project structure and code style.
velobase-harness copilot-instructions.md
Instructions for velobase/velobase-harness: Read AGENTS.md first. It is the English canonical AI agent guide for this repository.