The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation contains lists of injection patterns such as 'ignore previous instructions'. Analysis confirms these are part of a defensive regex filter implementation intended to protect the AI agent from external malicious emails, rather than an attempt to override the agent's behavior.
[PROMPT_INJECTION]: While the skill facilitates the ingestion of untrusted external content (emails), it proactively addresses indirect prompt injection risks. It provides exhaustive guidance on mitigation, including strict allowlists, content sanitization routines, and human-in-the-loop checkpoints to ensure safe processing of external data.
[EXTERNAL_DOWNLOADS]: The skill directs users to official SDKs from legitimate package registries (NPM, PyPI, etc.) and recommends well-known development tools like ngrok and Cloudflare Tunnel for secure local testing of webhook endpoints.

resend