What the skill teaches your agent
- Connect with no key. Add the MCP server at zambo.dev/mcp (remote streamable HTTP). No API key, no account, no credential handling of any kind.
- Work with a receipt on every call. The endpoint carries 153 tools, 35 on the default surface, and every call returns a verifiable receipt the agent keeps and reports with the result.
- Hand work over properly. Use the handover tools to pass the work to another AI or a client, with the state of the job and the receipt trail attached, so the receiver checks what ran instead of trusting a summary.
- Verify any receipt. At zambo.dev/verify in a browser, or with
GET https://zambo.dev/api/v2/receipt/{id}/verifyfrom code. - Respect the free tier. 20 free calls per day, then a Day Pass at $0.99 USDC on Base for 24h, or the Zambo Pass at $29 for 30 days. The skill tells the agent to say so before running past the free tier.
One definition the skill repeats word for word, because it is the honest scope of a receipt: an agent receipt is a record of one ai agent tool call that anyone can check: it keeps the exact saved bytes and their sha-256 digest, and it proves the result was not changed since it was saved, not that it was right.
What is in the zip
zambo-handover/
SKILL.md
scripts/
verify_receipt.pySKILL.md follows the Agent Skills specification at agentskills.io: the folder name matches the frontmatter name field (zambo-handover), and the frontmatter carries a description that tells an agent when to use it. verify_receipt.py is a short, auditable Python 3 helper (standard library only) that takes one receipt id and prints the public verify result. Read both files before you install. They are short on purpose.
Install, per client
Every route is the same shape: unzip the bundle into the client's skills folder, so the result is a zambo-handover folder holding SKILL.md, then start a new session. Paths below are personal (all projects) installs.
| Client | Skills folder | Command |
|---|---|---|
| Claude Code | ~/.claude/skills/ | unzip zambo-handover.zip -d ~/.claude/skills/ |
| Codex | ~/.agents/skills/ (some builds also read ~/.codex/skills/) | unzip zambo-handover.zip -d ~/.agents/skills/ |
| Cursor | ~/.cursor/skills/ | unzip zambo-handover.zip -d ~/.cursor/skills/ |
| Gemini CLI | ~/.gemini/skills/ | unzip zambo-handover.zip -d ~/.gemini/skills/ |
| GitHub Copilot | ~/.copilot/skills/ personal, or .github/skills/ inside one repo | unzip zambo-handover.zip -d ~/.copilot/skills/ |
| Goose | ~/.config/goose/skills/ | unzip zambo-handover.zip -d ~/.config/goose/skills/ |
| OpenClaw | ~/.openclaw/skills/, or the skills/ folder in your workspace | unzip zambo-handover.zip -d ~/.openclaw/skills/ |
Two honesty notes on that table. The Claude Code, Gemini CLI, and GitHub Copilot paths are stated in those projects' own skill documentation. The Goose and OpenClaw paths are the paths reported by current public skill indexes for those clients and were not confirmed in first party docs at publish time, so check your client's skills documentation if the folder differs on your version. For a single project instead of all projects, most clients also read a project local skills folder (for example .claude/skills/, .agents/skills/, or .github/skills/ in the repo root): unzip there and commit the folder if the team should share it.
This bundle is distributed only as this direct download from this hub. It is not listed on any third party skill marketplace, by design: a plain file you can read end to end before your agent ever loads it.
First run
After installing, ask your agent for work that ends in proof, for example: "Use zambo-handover: research this topic through Zambo, then hand the result over with the receipt links attached." The agent should connect with no key request, keep receipt links as it works, and offer the verifier link for anything it reports. If it ever asks you for a Zambo key or password, stop: the skill says never, and so does the endpoint.