The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill reads external user data which serves as a potential surface for indirect instructions.
Ingestion points: Context is gathered by reading goals.md in Step 1.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present when reading the file.
Capability inventory: Local file and directory creation within 04-workspace/ and updating the user-config.yaml configuration file.
Sanitization: No explicit sanitization of the content from goals.md is performed before processing.
Command Execution (SAFE): The skill's file system operations are limited to creating directories and markdown documentation within the designated workspace folder, consistent with its stated purpose.
Persistence Mechanisms (SAFE): The update to user-config.yaml is for state tracking and does not involve installing services or modifying shell profiles.
Data Exposure (SAFE): No network operations or hardcoded credentials were detected; the skill operates exclusively on local files.

setup-workspace