niche-research
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses explicit instructions to override the agent's standard conversational behavior, such as 'CRITICAL: Auto-start on load' and 'Do not summarise the research method.' These are low-severity attempts to control agent flow.
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by design. It is instructed to navigate to and process data from untrusted external platforms where malicious actors could embed instructions.
- Ingestion points: Browser-based scanning of Reddit feeds, X (Twitter) 'For You' feeds, and top Google search results.
- Boundary markers: The instructions lack explicit delimiters or 'ignore' directives to prevent the agent from potentially obeying commands found within social media posts or news articles.
- Capability inventory: The agent has the ability to read local project files (e.g., about-me.md), prompt the user for input, and utilize powerful live browsing tools (Claude for Chrome, Playwright, or WebSearch).
- Sanitization: No sanitization or filtering logic is specified for the content retrieved from external sites before it is synthesized into themes and presented to the user.
Audit Metadata