The Agent Skills Directory

Command Argument Interpolation: The skill constructs shell commands using user-supplied inputs for email addresses, role names, and environment URLs. This pattern represents a surface for indirect prompt injection where a user might attempt to inject additional command arguments or operators, although the skill's instructions to preview actions in plain prose act as a helpful human-in-the-loop checkpoint.
Ingestion points: User input for <email-or-object-id>, <role>, and <url> (SKILL.md).
Boundary markers: None identified.
Capability inventory: Role management and environment administration via pac CLI; API access via dataverse CLI; execution of local Python scripts for authentication checks.
Sanitization: No explicit sanitization or escaping logic is defined within the skill instructions for the interpolated shell arguments.
Privilege Management and Elevation: The primary function of this skill is to manage high-privilege roles (e.g., "System Administrator") and perform administrative self-elevation. To address the inherent risk of these operations, the skill defines a specific "Self-elevation confirmation protocol" that requires the agent to state risks explicitly, capture a justification (such as a ticket ID), and wait for a second confirmation after these details are presented.
Verification Logic for Security Actions: The skill identifies that certain CLI tools (PAC CLI) may return a successful exit code even if an operation fails due to invalid parameters. It instructs the agent to perform secondary verification via API queries to confirm that privilege changes were correctly applied, which is a positive security practice for ensuring the integrity of access control changes.

dv-security