The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly instructs the AI to ask the user for their PostHog community credentials (email and password) and provides a command to export them as environment variables (SQUEAK_EMAIL, SQUEAK_PASSWORD) in plaintext.
[DATA_EXFILTRATION]: The skill uses curl to send the collected credentials to an external Strapi backend (better-animal-d658c56969.strapiapp.com) to obtain a JWT. Although the skill author is 'posthog', this pattern of requesting and transmitting raw credentials via shell scripts is a high-risk data handling practice.
[COMMAND_EXECUTION]: The skill executes multiple complex shell scripts involving curl, python3 -c for JSON parsing, and temporary file creation (mktemp). These scripts process variables derived from the Substack URL and user input, increasing the surface area for command injection.
[DYNAMIC_CONTEXT_INJECTION]: The skill suggests the user run a shell command starting with ! (! export SQUEAK_EMAIL=...), which is the syntax for immediate shell execution in some agent environments. This bypasses typical prompt guardrails and executes the command directly in the host shell.

post-newsletter