fireflies-core-workflow-b
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external meeting transcripts and AI-generated responses from the AskFred service.
- Ingestion points: Data enters the agent's context through the
firefliesQuerywrapper used inSKILL.mdto fetch transcript summaries, keywords, and AI responses. - Boundary markers: The instructions explicitly remind the user to confirm raw transcript data is not sent to destinations unless permitted by access policies, establishing a manual check for data handling.
- Capability inventory: The skill utilizes GraphQL queries and mutations to retrieve data and interact with an external AI assistant. It uses
firefliesQueryfor network communication with the service provider. - Sanitization: The implementation focuses on structured data (JSON) and aggregation (e.g., keyword counting, duration summing), which naturally limits the impact of potential instructions embedded in transcripts.
Audit Metadata