The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external JSON command plans (craft-command-plan/v1) and project files (.pcraft), which inherently represents an attack surface for indirect prompt injection. However, this is the primary intended function of the tool. The skill implements extensive mitigations including strict schema validation, project hash verification (expectedProjectSha256), and path sanitization to ensure that operations remain within authorized boundaries.
[EXTERNAL_DOWNLOADS]: The skill uses npx skills add to fetch and install related components such as photocraft-harness from the full-aigc-skills organization. This is a standard and documented procedure within the platform's ecosystem for managing modular skill dependencies from a known author.
[COMMAND_EXECUTION]: The skill executes a native command-line interface (photocraft-cli) using Python wrapper scripts (workflow.py, commands.py). These executions are carefully managed, using the -I -B flags for the Python interpreter to isolate the environment and relying on a fixed catalog of commands (commands.json) to prevent the execution of arbitrary or unassigned operations.

photocraft-cli-layers