The Agent Skills Directory

[DATA_EXFILTRATION]: The skill reads local project files to extract architectural patterns and conventions. No instructions for network transmission, external API requests, or data harvesting to remote servers were found.
[COMMAND_EXECUTION]: The skill includes instructions to execute the project's existing type-check command to verify the validity of generated files. This is a localized verification step consistent with the tool's primary purpose as a developer aid.
[PROMPT_INJECTION]: The skill processes existing codebase files (exemplars) to guide its generation logic, which creates a potential surface for indirect prompt injection.
Ingestion points: SKILL.md (Step 2: FIND EXEMPLARS) reads 2-3 files from the user's project.
Boundary markers: Absent; the skill does not explicitly instruct the agent to ignore instructions embedded within the exemplars.
Capability inventory: File creation/modification (Step 4/5) and command execution (Step 6).
Sanitization: Absent; patterns are replicated directly from identified exemplars. This surface is inherent to project-aware scaffolding tools and is addressed via the localized development context.
[SAFE]: No malicious overrides, credential exposure, or persistence mechanisms were detected. The skill follows best practices such as confirming before overwriting existing files.

scaffold