The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external data (photos and text), which serves as an indirect prompt injection surface. However, the instructions incorporate strong safety markers, directing the agent to ignore privacy-sensitive metadata and to sanitize all user-provided identifiers into a safe format.
[SAFE]: The skill implements proactive file system security by instructing the agent to validate all output paths. It explicitly forbids the use of absolute paths or parent directory sequences (..) to prevent path traversal attacks outside the designated storage directory.
[SAFE]: Privacy protection is a core component of the workflow. The agent is strictly prohibited from storing, copying, or logging the local paths of original user photos, ensuring that only generated, non-private assets are included in the final package.

jinchenma-ip-builder