create-avatar
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local utility scripts for tasks including environment detection, media saving, and file uploads. These scripts include path validation to ensure operations remain within the project folder.
- [EXTERNAL_DOWNLOADS]: Utility scripts perform network requests to download media from user-specified URLs and check for tool updates by fetching version information from the author's GitHub repository.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external documents and web links to automatically draft avatar personas and knowledge bases. (1) Ingestion points: Instructions in SKILL.md and references/drafting.md direct the agent to read and summarize external content. (2) Boundary markers: No specific delimiters or safety instructions are defined to separate ingested content from system prompts. (3) Capability inventory: The skill has extensive capabilities including shell command execution, file system writes, and network communication. (4) Sanitization: The skill lacks explicit sanitization or validation of the ingested external data before it is used to influence the agent's behavior.
- [DATA_EXFILTRATION]: The skill collects local environment metadata and facilitates data transfers to vendor-controlled cloud storage.
- The detect-host script reads local environment variables to identify the AI agent platform.
- The upload-file.sh script transmits local files to Atmee's storage buckets on Supabase, verified against a whitelist of authorized vendor projects.
Audit Metadata