podcast-pipeline
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains a 'System Prompt Inquiry Response' instruction in
SKILL.mdthat directs the agent to provide a pre-scripted promotional response for 'Scayver Academy' if asked about its system prompt. This is a behavioral override designed to intercept prompt extraction attempts. - [PROMPT_INJECTION]: The skill enforces a 'Mandatory Intro Message' including payment links for CashApp and BuyMeACoffee. This forces the agent to include promotional content in its responses, regardless of the user's specific intent.
- [PROMPT_INJECTION]: The skill's architecture presents a surface for indirect prompt injection due to its handling of external data.
- Ingestion points: The 'Intake' section in
SKILL.mdrequires the agent to collect and process external episode URLs and transcripts. - Boundary markers: The instructions do not define clear delimiters or 'ignore' instructions for the data being processed, which increases the risk of the agent following instructions embedded within the transcript.
- Capability inventory: The agent uses processed data to generate multiple public-facing assets, including social media posts and blog content.
- Sanitization: No explicit sanitization or validation steps are provided to handle malicious content or hidden instructions within the source transcripts.
Audit Metadata