product-page-marketing-sections
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted marketing strategies from users to generate component code and implementation plans. A maliciously crafted strategy could attempt to influence the agent's code generation or subsequent workflow steps.
- Ingestion points: User-provided strategy text (prose, bullets, or files) and local project repository metadata.
- Boundary markers: The skill provides templates for internal notes to structure the implementation but does not mandate specific delimiters or escaping for the raw user-supplied strings.
- Capability inventory: The skill is authorized to write component files to the repository and execute local project-defined scripts (lint, build, test).
- Sanitization: The instructions contain strong counter-injection guidance, directing the agent to use only authentic evidence and verified facts while preserving existing commerce logic.
- [COMMAND_EXECUTION]: The workflow requires the agent to execute shell commands to verify the project's state and build success.
- Evidence:
SKILL.mdandreferences/integration-and-qa.mdinstruct the agent to run "applicable project checks (typecheck, lint, build, or targeted tests)" using the repository's own versioned scripts.
Audit Metadata