The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources (webhook providers like Stripe, GitHub, and Slack).
Ingestion points: Inbound HTTP webhook requests from third-party providers as described in SKILL.md and references/providers.md.
Boundary markers: Implements mandatory signature verification (verify-webhook) at the ngrok edge, creating a security boundary that validates source authenticity before delivery.
Capability inventory: The skill enables routing of verified payloads to internal services using forward-internal and involves secret management via the ngrok API.
Sanitization: Cryptographic signature verification is enforced; however, specific content-level sanitization of the webhook body (e.g., for downstream LLM consumers) is not explicitly detailed in the provided snippets.
[COMMAND_EXECUTION]: References the use of the ngrok CLI tool for managing secrets (ngrok api secrets create) and endpoints. These are standard vendor-provided administrative commands used for their intended connectivity purposes.
[SAFE]: The skill promotes several security best practices: it warns against hardcoding signing secrets, explicitly instructs agents not to request or accept secrets via chat, and advocates for the use of secure vaults over plaintext storage.

receive-webhooks