dify
Pass
Audited by Gen Agent Trust Hub on Mar 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Employs
bash -cto execute command strings that incorporate pipes and environment variables. This pattern is used to ensure theDIFY_TOKENsecret is correctly passed to thecurlcommand within the shell environment. - [DATA_EXFILTRATION]: Performs network requests to
api.dify.ai, which is the official endpoint for the Dify platform (a well-known service). The skill transmits user queries, configuration data, and local files to the platform as part of its documented functionality. - [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing and displaying content generated by an external LLM platform.
- Ingestion points: Responses from the Dify API via
curl(SKILL.md). - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are included in the command examples.
- Capability inventory: The skill utilizes shell execution (
bash), network operations (curl), and temporary file system access (/tmp). - Sanitization: External API responses are parsed for structure using
jqbut are not sanitized for potential malicious instructions before being presented to the user or agent.
Audit Metadata