The Agent Skills Directory

[SAFE]: The skill implements a comprehensive integrity management system that enforces SHA-256 digest verification for all project briefs, assets, and deliverables, as detailed in references/workflow.md and references/project-package.md. This prevents the processing of tampered or unauthorized files.
[EXTERNAL_DOWNLOADS]: Referenced machine learning models and supplementary tools are fetched from well-known trusted services (Hugging Face) or the official vendor repository (full-aigc-skills). These downloads are integrated into the skill's setup process following secure installation patterns.
[INDIRECT_PROMPT_INJECTION]: As the skill ingests external project briefs and media assets, it possesses a standard attack surface for indirect prompt injection. However, the risk is effectively mitigated by the use of strict JSON schema validation and a required human-in-the-loop technical review status (review_ready) before task completion.
[COMMAND_EXECUTION]: Subprocess calls for domain-specific creative tools are managed through a centralized harness. The orchestration scripts are invoked using Python's isolated mode (-I -B) to prevent environmental poisoning and maintain execution security.

artcraft-cli-assets