batch-file-creation
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructs the agent to generate and execute Python (
scratch/scaffold.py) or Node.js scripts at runtime to handle multiple file operations simultaneously. - [COMMAND_EXECUTION]: Promotes the use of shell commands (e.g.,
node -e "...") to execute complex logic that modifies the filesystem in a single turn. - [PROMPT_INJECTION]: Includes a "MANDATORY" operational directive that instructs the agent to disregard standard, granular tool calls (sequential
write_to_file) in favor of the higher-risk script execution approach. - [INDIRECT_PROMPT_INJECTION]: Identifies a significant attack surface where untrusted data or user-provided strings interpolated into these generated scripts could lead to code injection or directory traversal if the agent does not implement strict sanitization.
- Ingestion points: Conversational context where users provide component names, boilerplate logic, or file paths.
- Boundary markers: None; the provided templates do not include delimiters or instructions to ignore embedded commands in the data being written.
- Capability inventory: The skill uses
Path.write_text,fs.writeFileSync, and shell execution to create and run arbitrary scripts. - Sanitization: None; the templates directly interpolate string contents into filesystem calls without escaping or validation.
Audit Metadata