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 skills add dovahkiin-v/letterbox --skill letterbox-setupgit clone --depth 1 https://github.com/dovahkiin-v/letterboxWrote 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/skills/dovahkiin-v/letterbox/letterbox-setup)<a href="https://agentmods.dev/skills/dovahkiin-v/letterbox/letterbox-setup"><img src="https://agentmods.dev/badge/skills/dovahkiin-v/letterbox/letterbox-setup/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/dovahkiin-v/letterbox/letterbox-setup"><img src="https://agentmods.dev/badge/skills/dovahkiin-v/letterbox/letterbox-setup.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00078 | $0.01802 |
| Opus 5 | $0.00039 | $0.00901 |
| Sonnet 5 | $0.00016 | $0.00360 |
| Haiku 4.5 | $0.00008 | $0.00180 |
Grade B, and why
letterbox-setup scanned grade B with 1 finding 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 9d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
1. Ensure `~/.gemini/settings.json` contains a `letterbox` MCP server under How it starts
The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Letterbox — setup & wiring guide
Letterbox lets two terminal coding agents hold a real-time conversation by
passing JSON message files through a shared directory (~/.letterbox/channels/<channel>/).
This skill covers the one-time wiring to make a harness letterbox-capable.
For using a live bridge (sending, reading, directed messages), see the separate
letterbox skill.
The shape of a bridge
A bridge is two launches — one per side, same channel:
letterbox <harness> --channel <name> --as <label>
letterbox <harness> spawns the harness, watches the channel, and injects 📬
notifications. The one prerequisite is that the target harness has the letterbox
MCP server registered so its agent gets the tools. Registration is one-time
per harness and differs by harness (below).
You cannot add MCP tools to your own already-running session. Registration
takes effect on the next launch of that harness. To bring up a peer: ensure
the wiring below exists for its harness, then spawn
letterbox <harness> --channel <name> --as <peer-label>.
Two values you'll reuse:
- the absolute path to the
letterboxbinary (readlink -f "$(command -v letterbox)"), - the launch directory must be trusted by Gemini/Antigravity (they otherwise pop a blocking "trust this folder?" prompt that stalls automation).
Identity: one label, one participant per channel
Each --as <label> must be unique on its channel. Launching a second
session under a label already running on that channel is refused:
'claude' is already running on channel 'demo' (pid 12345).
Use a different name to run a second instance, e.g.:
letterbox claude --channel demo --as claude-2
This is deliberate — it makes the label the atomic unit of identity, which is
what lets directed addressing (send_message(to="…")) be unambiguous. For
multiple agents of the same kind in one room, give them distinct labels:
claude-review / claude-commit, gemini-1 / gemini-2. The same label on a
different channel is fine (different conversation, no collision).
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.
- 9d ago First seen · 161 lines · 78 tokens per session scan B 2b05e71b6ea4
letterbox-setup is a skill published in the GitHub repository dovahkiin-v/letterbox (0 stars, last pushed 16d ago), licensed MIT. It adds 78 tokens to every session and 1,802 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
use-agent-supagents
Use when authoring or building supagents sources under .agents/supagents/ — markdown that compiles to subagent files for Claude Code, Gemini CLI, GitHub Copilot, Cursor, OpenCode, and Kilo Code.
project-adopt
Onboarding para projetos existentes que recebem a estrutura Claude Code pela primeira vez. Descobre convenções do codebase antes de perguntar qualquer coisa.
project-init
Onboarding da primeira sessão para projetos novos. Execute automaticamente quando MEMORY.md não tiver contexto. Entrevista o desenvolvedor e configura todos os arquivos de memória.
publish-pattern
Publica um padrão reutilizável no repositório pessoal de padrões do usuário, se houver um configurado. Use quando uma solução se provar genuinamente reutilizável em outros projetos.
bugfix
Protocolo sistemático de triage de bugs. Use ao receber report ou comportamento inesperado. Segue a sequência reproduzir→localizar→reduzir→corrigir→guardar→verificar.
project-seal
Fecha o setup inicial do projeto — commita specs, documentação e configuração após project-init + primeira spec. Roda uma vez, logo após a primeira spec ser criada e revisada.