brand-guidelines
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill name, description, and internal instructions repeatedly claim to provide "Anthropic's official brand colors," "official brand identity," and "company design standards." However, the skill is authored by 'aboudjem' rather than Anthropic. Furthermore, the provided
LICENSE.txtcontains a copyright notice attributed to "Anthropic, PBC," which constitutes deceptive metadata regarding the origin and official status of the skill. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external artifacts to apply visual styling, which creates a surface for indirect prompt injection where instructions embedded in the processed data could influence agent behavior.
- Ingestion points: Content within artifacts provided by the user (as described in
SKILL.md). - Boundary markers: None identified in the skill instructions to delimit untrusted content from the styling logic.
- Capability inventory: The skill description indicates it utilizes styling logic and the
python-pptxlibrary to modify presentation files and other artifacts. - Sanitization: There is no mention of sanitization, validation, or filtering of the input data before it is processed for styling.
Audit Metadata