The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the official @relaymessenger/sdk package from the NPM registry as described in references/sdk-and-auth.md. These dependencies are provided by the skill's authoring organization and are used for standard integration purposes.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes external data from webhooks and documentation tools.
Ingestion points: Data enters the context through the relayDocs MCP tool (defined in agents/openai.yaml) and via agent events such as Webhooks and WebSockets (described in references/agent-events.md).
Boundary markers: SKILL.md mandates the use of a "locked contract" approach, referencing references/relay-v1-lock.json to ensure the agent only uses verified API schemas.
Capability inventory: The skill includes functions for sending messages (references/messaging.md), managing webhook subscriptions (references/agent-events.md), and modifying contact cards (references/chats-and-contacts.md).
Sanitization: references/messaging.md provides a relayApiOrigin function to validate and sanitize API base URLs. Additionally, references/agent-events.md includes instructions to reject local, private, or metadata service addresses in webhook destinations to prevent SSRF attacks.

relay