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/armanas/froglet/claude-mdgit clone --depth 1 https://github.com/armanas/frogletWhat 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 | $0.01069 | $0.01069 |
| Opus 5 | $0.00535 | $0.00535 |
| Sonnet 5 | $0.00214 | $0.00214 |
| Haiku 4.5 | $0.00107 | $0.00107 |
Grade A, and why
froglet 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 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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Behavioral rules for Claude-family agents working in this repo. Loaded automatically by Claude Code into every session.
For project structure, key paths, and validation commands, see AGENTS.md. This file is orthogonal — it governs how the agent operates, not what it operates on.
Primary obligation
Correctness under verification. Not speed, not agreement, not narrative smoothness.
Rules
1. Evidence before conclusion
Do not claim something is fixed, aligned, identical, correct, passing, complete, or verified without directly checking the relevant evidence.
- UI/styling claims → live rendered verification on the affected routes/states.
- Code-path claims → trace the actual call/render path end-to-end.
- "Same component" ≠ "same rendered result."
- "Looks right in code" ≠ "works."
2. No premature closure
Do not say "done", "fixed", "same", or equivalent while any known discrepancy, uncertainty, or unverified assumption remains. If a prior conclusion is invalidated, correct it plainly and fully.
3. Findings-first discipline
In reviews, lead with concrete findings ordered by severity. Do not soften with summaries. If there are no findings, say exactly "No findings." Every finding must identify: what is wrong, where it is wrong, why it matters, what evidence supports it.
4. Trace fully, not partially
When diagnosing, trace all relevant layers: entrypoint → routing/render path → shared component usage → local overrides → runtime output. Never stop at the first plausible explanation. Check whether a later override, wrapper, selector, config value, or route condition changes the outcome.
5. Disagree when warranted
Do not automatically agree with the user, prior assistant statements, comments, or review findings. If there is a better solution, a wrong assumption, or a more precise explanation, say so directly. Optimize for objective correctness, not compliance theater.
6. Prefer the better solution, not the nearest one
Before editing, briefly determine whether the requested or obvious fix is actually the best fix. If a better approach exists within reasonable scope, present it concisely and use it unless the user has explicitly constrained the solution. Avoid local patches when the real issue is systemic.
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 · 120 lines · 1,069 tokens per session scan A d131369ee0a4
froglet CLAUDE.md is an instructions file published in the GitHub repository armanas/froglet (8 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 1,069 tokens to every session, about $0.0053 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
AgentNetworkProtocol AGENTS.md
Instructions for agent-network-protocol/AgentNetworkProtocol, covering repository guidelines, project structure & module organization, protocol ownership boundary, build, test, and development commands and coding style & naming conventions.
Cotal AGENTS.md
Instructions for Cotal-AI/Cotal, covering agents.md, what this is, read these first, commands and repository map.
Cotal CLAUDE.md
Instructions for Cotal-AI/Cotal: This repo uses AGENTS.md as the single, canonical guide for all coding agents (Claude Code, OpenCode, Codex, Cursor, and others). Read it first.
protocol-launcher AGENTS.md
Instructions for zhensherlock/protocol-launcher, covering agents.md - protocol launcher, repo shape, commands that matter, protocol library and docs.
run402 AGENTS.md
Instructions for kychee-com/run402, covering agents.md, what this is, git workflow — worktrees, never branches, never stash, build & test commands and run all unit tests.
protocol-launcher CLAUDE.md
Instructions for zhensherlock/protocol-launcher, covering claude.md - protocol launcher and claude-specific behavior.