retrodiffusion
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for environment management and to run the generation script. It uses
test,grep, andsource(. .env) to handle theRETRODIFFUSION_API_KEY. Sourcing a.envfile that contains unsanitized user input poses a risk of arbitrary command execution. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential command injection surface because user-provided content (the
[prompt]argument) is interpolated into a shell command line (node scripts/retrodiffusion-generate.mjs ... --prompt "[prompt]"). - Ingestion points: The user-supplied
[prompt]value in theSKILL.mdinstructions and potentially the API key input. - Boundary markers: No specific delimiters or instructions to ignore embedded commands are present for the user prompt.
- Capability inventory: The skill executes shell commands and writes to the local file system (asset folders and the .env file).
- Sanitization: No sanitization logic or validation for user input is described before it is used in the command execution string.
Audit Metadata