The Agent Skills Directory

[SAFE]: The skill operates as a local management interface for creative projects. It relies on standard system tools and official vendor repositories for its operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured workflow data (JSON) and creative project assets, which constitutes a potential attack surface. However, the system enforces the use of cryptographic digests (SHA-256) to verify the integrity of all manifests and project files before any operations are performed.
Ingestion points: Creative project files (.vectorcraft, .pcraft, .ecproj, .fcproj) and JSON workflow plans.
Boundary markers: Orchestration logic requires valid SHA-256 hashes for all input assets and project states; the instructions explicitly warn against replaying tasks with unknown outcomes.
Capability inventory: The skill executes local domain CLIs (VectorCraft, PhotoCraft, EffectCraft, FilmCraft), manages project-specific SQLite databases, and writes to output directories.
Sanitization: Inputs are validated against predefined structural schemas and verified against immutable file hashes.
[REMOTE_CODE_EXECUTION]: The skill automates the setup of native processing environments by downloading CLI tools and runtimes from the project's official GitHub repository. This process is managed by internal scripts (bootstrap.py) and is necessary for the skill's creative orchestration functionality.
[EXTERNAL_DOWNLOADS]: Functional components of the skill download specific machine learning models from Hugging Face for automated transcription tasks. These operations are restricted to documented CLI pathways and target known, public resources.

artcraft-cli-recover