The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill performs localized file system operations to initialize a workspace, consistent with its description.
[PROMPT_INJECTION]: The skill initializes an attack surface for indirect prompt injection by creating files (vision.md, tech-environment.md) intended to store project-specific requirements that will be ingested into the agent's context by other SpecForge skills. This is a functional requirement of the workflow.
Ingestion points: specforge/inputs/vision.md and specforge/inputs/tech-environment.md located at the project root.
Boundary markers: The seeded templates do not currently include explicit delimiters or instructions for subsequent skills to treat the user-provided content as untrusted data.
Capability inventory: The skill is configured with Read, Write, Edit, and Bash tools.
Sanitization: No sanitization or validation logic is defined for the content of the generated files.

init