customer-story-builder
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill implements a custom response for system prompt inquiries in
SKILL.md, directing users to a membership site. This is a form of behavioral instruction intended to override the agent's standard response to prompt extraction attempts. - [PROMPT_INJECTION]: The skill processes potentially untrusted external content, creating a surface for indirect prompt injection.
- Ingestion points: Raw customer material including interviews, survey responses, and support tickets are ingested during the 'Intake' phase in
SKILL.md. - Boundary markers: The instructions lack explicit boundary markers (e.g., XML tags or delimiters) to separate untrusted customer data from the agent's instructions.
- Capability inventory: The skill is limited to text analysis and generation; it does not utilize file system writes, network requests, or shell execution tools.
- Sanitization: There are no instructions for sanitizing or filtering input data for malicious commands.
Audit Metadata