infographic
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONREMOTE_CODE_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted user text (as described in SKILL.md) and interpolates it into JavaScript code that is subsequently executed by the use_figma tool. There are no boundary markers or sanitization steps defined in the skill to prevent malicious input from potentially breaking the generated script or interfering with the execution environment.
- [COMMAND_EXECUTION]: The skill uses a curl command to upload a local brand avatar image to a dynamic URL destination provided by the agent's upload tool.
- [DATA_EXFILTRATION]: The skill reads a local file path (~/.claude/skills/infographic/assets/avatar.png) and transmits the file content to a tool-specified network endpoint. While restricted to brand assets, this pattern involves the reading and external transmission of local data.
- [REMOTE_CODE_EXECUTION]: The skill performs dynamic JavaScript generation and execution via the use_figma tool to construct and manipulate design elements within the Figma design environment.
Audit Metadata