boost-prompt
Warn
Audited by Gen Agent Trust Hub on Apr 14, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a Clojure code snippet to be executed through the Joyride VS Code extension. This allows the agent to perform actions like modifying the system clipboard using
(vscode/env.clipboard.writeText ...). - [REMOTE_CODE_EXECUTION]: The agent is instructed to dynamically generate and execute code at runtime. This execution pattern is risky as it creates a surface where an agent might be tricked into running unauthorized commands if the data being processed is not strictly validated.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it performs 'project explorations' to understand tasks. This process involves reading untrusted files from the project environment which are then used to build the final prompt and the executable script.
- Ingestion points: Project files read during the exploration phase in SKILL.md.
- Boundary markers: Absent; there are no delimiters or instructions provided to the agent to treat external project content as untrusted.
- Capability inventory: Execution of Clojure scripts via Joyride; project file reading via exploration tools.
- Sanitization: Absent; there is no logic to escape or sanitize the content gathered from the project before inserting it into the prompt or the executable Clojure string.
Audit Metadata