The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes data from external Okta environments that could contain malicious payloads designed to manipulate agent behavior.
Ingestion points: Data retrieved from Okta actions such as list-users, list-groups, and get-application (SKILL.md).
Boundary markers: The skill includes a 'Security Guardrails' section with explicit natural language instructions to prevent exfiltration and require confirmation for destructive tasks (SKILL.md).
Capability inventory: The skill possesses destructive capabilities including delete-user, delete-group, and the ability to make arbitrary API requests via membrane request (SKILL.md).
Sanitization: The skill relies on prompt-level instructions rather than programmatic sanitization of the external API data.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the @membranehq/cli package globally via npm (npm install -g @membranehq/cli), which is the official tool provided by the vendor for this integration (SKILL.md).
[COMMAND_EXECUTION]: The skill documentation provides instructions for the agent to execute shell commands using the membrane CLI to manage authentication and interact with Okta resources (SKILL.md).

okta-hardened