gds-brainstorm-game
Fail
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: HIGHDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The SKILL.md file contains an On Complete section that instructs the agent to execute a command string resolved from the workflow.on_complete configuration key. This enables arbitrary shell command execution at the end of a session, controlled by configuration files located in the project root.
- [COMMAND_EXECUTION]: The workflow relies on uv run to execute local Python scripts from the project directory. This establishes a dependency on the integrity of the project environment and allows for potential execution of malicious scripts if the project directory is compromised.
- [INDIRECT_PROMPT_INJECTION]: 1. Ingestion points: The skill merges configuration from .toml files in the project directory and captures user-provided brainstorming ideas. 2. Boundary markers: No delimiters or warnings are used. 3. Capability inventory: The agent has permissions to execute shell commands and write to the filesystem. 4. Sanitization: The skill lacks explicit sanitization or validation logic for data ingested from the filesystem or user dialogue.
Recommendations
- AI detected serious security threats
Audit Metadata