The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents HTTP action handlers that ingest data from external sources such as Stripe webhooks and generic API requests. The examples promote secure practices by including signature verification (verifyStripeSignature) and JSON schema validation using the convex/values library, which significantly mitigates the risk of processing malicious inputs.
[EXTERNAL_DOWNLOADS]: The skill references several documentation URLs hosted on docs.convex.dev and convex.dev to provide context for the AI agent. These are official documentation sites for the platform described in the skill.
[COMMAND_EXECUTION]: The Best Practices section includes explicit negative constraints, instructing the agent to avoid running sensitive commands like npx convex deploy or git operations unless explicitly directed by the user, which acts as a safety guardrail.
[DATA_EXFILTRATION]: A code example for a Convex Action demonstrates the use of the fetch API to interact with an external service (api.email.com). This is presented as a generic integration template and does not involve the exfiltration of sensitive system credentials or files.

convex-functions