research-prompt
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill generates research prompts by interpolating data from external sources, creating a vulnerability surface where ingested instructions could override the skill's logic.
- Ingestion points: Context is retrieved from "relevant project files" and the "conversation" history during the generation process.
- Boundary markers: The template and instructions do not include delimiters or specific instructions for the agent to treat ingested context as untrusted data or to ignore embedded commands.
- Capability inventory: The resulting prompt is intended for execution via "DeepAPI POST /v1/research/deep" or other deep-research agents, which typically have network access and data retrieval capabilities.
- Sanitization: There are no instructions or mechanisms to sanitize, filter, or escape characters from the ingested context before it is embedded in the final output paragraph.
Audit Metadata