The Agent Skills Directory

Indirect Prompt Injection (SAFE): The skill implements endpoints to ingest external data from SendGrid webhooks.
Ingestion points: POST /webhooks/sendgrid in examples/express/src/index.js, examples/fastapi/main.py, and examples/nextjs/app/webhooks/sendgrid/route.ts.
Boundary markers: The code verifies the cryptographic signature (ECDSA) provided by SendGrid before processing the payload, which ensures the data originates from a trusted source.
Capability inventory: The handlers are limited to parsing JSON and logging event details to the console. No high-privilege operations or network requests are triggered by the payload content.
Sanitization: Input is parsed as JSON. While content is logged raw, the signature verification acts as a strong integrity control.
Metadata Poisoning (SAFE): The metadata accurately describes the skill's purpose. Version numbers in dependency files (e.g., next@16.1.6, pytest@9.0.2) appear to be slightly ahead of current stable releases, likely due to future-dated boilerplate or template generation, but no malicious dependency patterns were detected.
Credentials Safety (SAFE): The skill contains example public keys and a test private key in the test suites (test_webhook.py, webhook.test.js). These are standard for unit testing cryptographic logic and do not represent a leak of production credentials.

sendgrid-webhooks