The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the lightcraft-cli and the main LightCraft binary to perform photo processing tasks. It executes subcommands such as run, render, snapshot, and merge to automate the application's engine.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of various image formats (JPEG, CR3, RAW, etc.) and XMP sidecar files, which can contain user-controlled metadata.\n
Ingestion points: The skill imports photos and reads XMP sidecars (crates/engine/src/import.rs, crates/engine/src/sidecar.rs) to retrieve ratings, flags, and edit settings.\n
Boundary markers: There are no specific instructions or delimiters mentioned to prevent the agent from interpreting instructions potentially embedded within image metadata fields.\n
Capability inventory: The agent has the capability to write files (exports), execute local CLI tools, and communicate with a local server on port 7980.\n
Sanitization: The analysis did not find evidence of sanitization or filtering of external metadata content before it is processed by the agent's context.\n- [EXTERNAL_DOWNLOADS]: The skill documentation mentions the ability to download AI models for masking tasks (e.g., SAM 3 for Object/Describe masks) and the requirement for users to manually install weights for AI Denoising. These downloads are directed to the project's own infrastructure or established repositories.

lightcraft