positioning-angles
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection through its competitive web search functionality.\n
- Ingestion points: In Step 2.5 (Competitive web search), the skill fetches headlines, hero copy, and CTAs from external competitor URLs.\n
- Boundary markers: None are specified; the skill does not explicitly instruct the agent to ignore or delimit embedded instructions within the crawled web content.\n
- Capability inventory: The agent possesses file-writing capabilities to the
./brand/directory and can perform further tool-based web searches.\n - Sanitization: No validation or sanitization of the retrieved web data is mandated before the agent processes it for strategy generation.\n- [COMMAND_EXECUTION]: The skill performs file system operations by reading several brand-related files (e.g.,
voice-profile.md,audience.md) and writing a finalizedpositioning.mdfile to the project's./brand/directory. These operations are transparently defined as part of the skill's business logic and are restricted to the project root.\n- [DATA_EXFILTRATION]: The skill retrieves internal brand context, including audience profiles and competitor lists, to provide personalized output. This data is included in the prompt context sent to the LLM. While this is the intended data flow for marketing analysis, it represents a risk factor if the project context contains sensitive but non-credential information.
Audit Metadata