Wappy handles real credentials: WhatsApp Cloud API access tokens, model provider API keys, and webhook signing secrets. Please report security issues responsibly rather than opening a public issue.

Email csrj0425@gmail.com with:

• A description of the issue and its potential impact.
• Steps to reproduce, or a proof of concept if you have one.
• Which package(s)/version(s) are affected.

You should get an acknowledgment within a few days. Please don't publicly disclose the issue until a fix has shipped, or 90 days have passed, whichever comes first.

In scope:

• @wappy_ai/core, @wappy_ai/harness, @wappy_ai/whatsapp, @wappy_ai/create-agent (this repo).
• Anything that could leak a credential, bypass webhook signature verification, allow a confirm-before-write tool to execute without real confirmation, or allow a crafted inbound message to execute unintended code.

Out of scope:

• Vulnerabilities in a generated project's own custom tools/connectors, unless the root cause is in this repo's own code.
• The WhatsApp Cloud API or any model provider's own infrastructure (report those to Meta/the provider directly).

A few things worth knowing up front, since they shape what a report should focus on:

• Secrets are never generated into committed files. .env/.wappy/ are git-ignored by every generated project; credentials only ever live in environment variables, read at call time.
• Every outbound HTTP call this repo makes goes through an SSRF guard (address-classification allow-list, DNS-rebinding-safe connection pinning, redirect re-validation on every hop).
• Webhook signatures are verified before a payload is ever processed.
• A tool marked confirmBefore: true is held pending, keyed to the specific confirmation it was issued for, and can't be replayed or confused with a different pending action (see packages/harness/src/confirm.ts).