viral-app
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from social media platforms (video titles, hashtags, and descriptions) through the viral.app API. Because this data is provided by external users and ingested by the agent to generate reports, it constitutes a surface for indirect prompt injection. Malicious instructions could potentially be embedded in video metadata to influence the agent, which has the capability to perform mutations such as initiating payouts.\n
- Ingestion points: Retrieves video and account metadata via CLI commands like
videos-listandanalytics-top-videosdescribed in SKILL.md.\n - Boundary markers: Templates in
assets/templates/lack explicit delimiters or instructions for the agent to ignore content-based commands.\n - Capability inventory: The skill can execute sensitive actions such as
payouts-initiateandprojects-createusing theviral-appCLI.\n - Sanitization: No specific sanitization or filtering of platform-retrieved text is mandated in the instructions or scripts.\n- [COMMAND_EXECUTION]: The skill relies on a vendor-provided CLI tool
viral-appto interact with the viral.app API and perform all operations, including reading analytics and initiating payouts.
Audit Metadata