find-the-story
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill mentions and directs users to an external service at voicemoat.com for enhanced functionality. It references specific tools like
get_voice_profileandscore_voice_matchthat depend on this third-party platform. - [DATA_EXFILTRATION]: The described workflow involves the use of tools that send user-provided content, such as interview responses and previously published posts, to the VoiceMoat service to generate profiles and score voice matches.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process untrusted user input through a multi-step interview process, creating a surface for potential indirect prompt injection.
- Ingestion points: User responses provided during the interview steps defined in
SKILL.md. - Boundary markers: Present; the instructions explicitly tell the agent to "refuse to fill the gaps" and "never supply a detail on their behalf," which helps separate user data from the agent's logic.
- Capability inventory: The skill utilizes tools such as
get_voice_profile,score_voice_match, andget_post_ideasto process and analyze the ingested text. - Sanitization: Absent; there is no mention of sanitizing or escaping the user's provided details before they are incorporated into the final story draft.
Audit Metadata