The Agent Skills Directory

[SAFE]: The skill consists of documentation and design guidelines for developers. It does not contain executable code or malicious instructions targeting the agent or environment.
[CREDENTIALS_UNSAFE]: The skill provides instructions for managing Apify API tokens securely, emphasizing best practices to avoid exposure.
Evidence: It explicitly warns against hardcoding tokens or logging them in chat logs or command output.
Evidence: It recommends storing tokens in environment variables (APIFY_TOKEN) or secret managers.
Evidence: It suggests using SecretStr in Python for Pydantic models to ensure tokens are automatically excluded from logs and serialized representations.
[INDIRECT_PROMPT_INJECTION]: The skill identifies and provides mitigation strategies for the risk of indirect prompt injection when agents process external data scraped by Actors.
Evidence: Recommends using explicit boundary markers such as <<<EXTERNAL_UNTRUSTED_CONTENT>>> to delimit untrusted data.
Evidence: Instructs developers to sanitize markers within the scraped data and cap payload sizes to prevent context blowout or prompt hijacking.
[EXTERNAL_DOWNLOADS]: The skill references official Apify documentation, APIs, and libraries, which are recognized as safe vendor resources.
Evidence: URLs target apify.com, console.apify.com, and api.apify.com for documentation and API access.
Evidence: Library references include apify-client (JS and Python) and official plugins like @apify/n8n-nodes-apify and langchain-apify.
[COMMAND_EXECUTION]: The skill describes the use of standard CLI tools for development and deployment workflows.
Evidence: Instructions mention using apify-cli for creating, running locally, and pushing Actors to the Apify platform.

apify-integration-development