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/harebeats/cannonball/agents-mdgit clone --depth 1 https://github.com/harebeats/cannonballWhat 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 | $0.00596 | $0.00596 |
| Opus 5 | $0.00298 | $0.00298 |
| Sonnet 5 | $0.00119 | $0.00119 |
| Haiku 4.5 | $0.00060 | $0.00060 |
Grade A, and why
cannonball 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- cannonball CLAUDE.md — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 51 lines — stays where its author put it; the contents beside it link to each section on GitHub.
cannonball — para qualquer agente
Este repositório é um conjunto de Agent Skills mais o motor em Python que elas chamam. Ele não depende de nenhum agente específico.
O que você precisa saber para operar
Cada pasta em skills/ é autocontida. Ela traz o SKILL.md, o scripts/ que ele
chama, o references/ que ele cita e o seed/ das peças de exemplo. Instalar uma
skill é copiar a pasta dela; nada fica para trás.
Resolva o caminho pelo arquivo, não por variável. Todo comando nas skills usa
${SKILL_DIR}, que é a pasta do SKILL.md que você acabou de ler. Nenhuma variável
de ambiente de agente é necessária — e é isso que faz o mesmo SKILL.md funcionar em
Claude Code, Codex, Gemini CLI, Cursor e no resto.
Requisito único: Python 3. Os scripts importam só a biblioteca padrão. Sem
pip install, sem node_modules, sem rede — exceto onde o próprio comando diz que
usa rede (curar.py --assets testa URLs).
Onde ficam os dados
O acervo é do usuário e mora fora deste repositório. A resolução, em ordem:
- variável
CANNONBALL_ACERVO cannonball.config.jsonna raiz, se a pasta existir~/.cannonball/aonde— ponteiro escrito porscripts/vincular.py../acervorelativo aos scripts, se você estiver rodando de um clone com dados~/.cannonball— padrão, criado na hora com as peças de exemplo
Nunca grave acervo dentro da pasta da skill instalada: em vários hospedeiros ela é uma cópia em cache que a próxima atualização apaga.
A ordem que dá melhor resultado
vincular.py --para <pasta> → kit-ingerir → kit-buscar → kit-cor + kit-tipo
→ kit-montar → kit-otimizar-3d → kit-ingerir (guarda o que deu certo)
python scripts/perfil.py imprime essa ordem já adaptada ao estado do acervo.
Ferramentas de agente citadas
As skills mencionam múltipla escolha para perguntas fechadas. Se o seu agente tiver uma ferramenta dessas, use; se não, pergunte em texto — nada quebra.
Quatro MCPs são opcionais e cada skill só dispara a seção correspondente quando o servidor está ligado: GetLayers, Motion Sites, OriginKit e Higgsfield. Nenhum é requisito.
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 · 51 lines · 596 tokens per session scan A ff32b5b9e53c
cannonball AGENTS.md is an instructions file published in the GitHub repository harebeats/cannonball (2 stars, last pushed 5d ago), licensed MIT. It adds 596 tokens to every session, about $0.0030 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
www-sacred AGENTS.md
Instructions for internet-development/www-sacred, covering agents.md, what this repo is, repo map, conventions and the fonts menu previews.
oh-my-design AGENTS.md
Instructions for kwakseongjae/oh-my-design, covering agents.md — oh-my-design, repository quick map, build / test / lint, design harness mode and entrypoint.
oh-my-design CLAUDE.md
Instructions for kwakseongjae/oh-my-design, a project described as: Give your AI coding agent a design system. One command installs 400+ quality-graded company DESIGN.md references + skills into Claude Code, Codex, Cursor, and OpenCode. Free, MIT, zero AI calls.
nextjs-admin-dashboard AGENTS.md
AGENTS.md instructions for NextAdminHQ/nextjs-admin-dashboard, covering nextadmin pro — ai agent rules, project structure, next.js, conventions and styling rules.
nextjs-admin-dashboard CLAUDE.md
Claude Code instructions for NextAdminHQ/nextjs-admin-dashboard, a project described as: AI-Native Next.js admin dashboard template and UI components that come with pre-built elements, components, pages, high-quality design, integrations, and much more.
daisyui code_generation_with_git.instructions.md
Instructions for saadeghi/daisyui, covering activation, initial setup, development loop, error handling and session end.