OTPBox / Skills

The OTPBox Agent Skill

A Skill is a packaged set of instructions that teaches a coding agent the right way to do something — not just which tools exist, but the workflow, the gotchas, and when to reach for it. The OTPBox Skill teaches an agent the full email-verification-testing loop: mint a key, create an inbox, block for the message, extract the code or link, clean up — and the mistakes to avoid (reusing an inbox across runs, polling instead of blocking, burning quota on unnecessary reads).

Install it

Save the skill file into your project's skills directory. For Claude Code:

mkdir -p .claude/skills/otpbox
curl -o .claude/skills/otpbox/SKILL.md https://otpbox.io/agent-skills/otpbox/SKILL.md

Any agent that reads Skills from a local directory (Claude Code and several others use a SKILL.md-in-a-folder convention) can use it the same way — drop it into that agent's own skills path.

What's in it

• When to reach for OTPBox — signup, login, 2FA, magic-link, and QA-data-generation tasks, and explicitly not agent-owned mailbox use cases.
• The core loop — create an inbox, use its address, block on the message, read the extracted code/link, clean up.
• Both call paths — MCP tools when the client supports it, REST endpoints when writing test code directly.
• Test identities — generating realistic personas instead of just addresses, single or in bulk.
• Gotchas — quota accounting per call, the 60-calls/minute MCP rate limit, the 25-second wait_for_email cap, and why sandbox keys never fire webhooks.

Why a Skill instead of just the MCP tools

The MCP server gives an agent the tools; the Skill gives it the judgment to use them well — when to call wait_for_email instead of polling, why an inbox should never be reused across test runs, and how to structure a bulk test-identity request instead of looping 50 single calls. An agent with the MCP server but not the Skill can still do the job, just less efficiently and with more trial and error against your quota.

Next steps