The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or security bypass instructions were detected in the skill content.
[COMMAND_EXECUTION]: The skill requests access to the Bash tool with curl capabilities. The provided instructions use this for legitimate interactions with Guidewire APIs as part of the intended RBAC implementation, with no evidence of command injection or malicious payload construction.
[INDIRECT_PROMPT_INJECTION]: The skill describes a system for processing external security tokens and user data.
Ingestion points: SKILL.md contains code snippets for extracting security context from JWT payloads and processing user roles.
Boundary markers: The provided templates include authorization middleware (requireApiRole) and validation logic to delimit untrusted input.
Capability inventory: The skill uses Bash(curl:*), Read, Write, Edit, and Grep tools.
Sanitization: The implementation patterns include audit logging (auditPermissionDenial) and structured permission verification logic to handle external inputs safely.

guidewire-enterprise-rbac