kit-onboarding
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation specifies that image and video generation should be routed through the
Higgsfield CLI. This involves the execution of external command-line tools for media creation. - [INDIRECT_PROMPT_INJECTION]: The skill includes a step (G2) where it ingests data from user-provided URLs to extract brand information. This represents an external data ingestion surface where malicious content on a website could attempt to influence agent behavior during the extraction process.
- Ingestion points: User-provided live site URLs (G2).
- Boundary markers: None explicitly defined for the extraction phase.
- Capability inventory: File creation/append (
marketing/DESIGN.md,CLAUDE.md), CLI execution (Higgsfield CLI). - Sanitization: None specified for the content extracted from external URLs.
- [PROMPT_INJECTION]: The skill modifies the agent's core instruction file (
CLAUDE.md) by appending new rules. While this is the intended purpose of the skill, modification of system-level instructions is a sensitive operation.
Audit Metadata