review-analyst-agent
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted content from external platforms such as Amazon, Reddit, and G2. Malicious reviews could be crafted to include instructions that attempt to bias the sentiment analysis or manipulate the synthesis of the final report.
- Ingestion points: Customer reviews are collected from various external sources in
SKILL.md(Step 2) andreview-scraper.md. - Boundary markers: The instructions do not specify the use of delimiters or specific warnings for the agent to disregard instructions embedded within the review text.
- Capability inventory: The skill possesses the ability to scrape web content, synthesize data into structured JSON, and execute local shell commands for PDF generation.
- Sanitization: There is no evidence of sanitization or filtering of the extracted review text before it is processed by the sentiment and issue analysis agents.
- [COMMAND_EXECUTION]: The skill includes a shell command in
SKILL.mdto execute a Python script (report_to_pdf.py) located in a sibling skill directory (media-utils). This is used for generating analysis reports and references a local environment path.
Audit Metadata