content-strategy
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a process that ingests data from untrusted external sources, specifically public forums (Reddit, Quora) and competitor websites, to generate content ideas and gaps.
- Ingestion points: Step 6 (Source Content Ideas) directs the agent to perform web searches and analyze content from Reddit threads, Quora questions, and competitor blogs.
- Boundary markers: The instructions do not define boundary markers or include specific directives to the model to treat content from these external sources as untrusted data or to ignore embedded instructions.
- Capability inventory: The skill utilizes web research and internet access to retrieve data from external domains and processes user-provided inputs like call transcripts.
- Sanitization: The process does not include steps for sanitizing, filtering, or validating the integrity of the external data before it is processed by the agent.
Audit Metadata