Image Editing

**Quick judgment** A generally strong, multi-purpose AI image editing skill that provides structured workflows for most common editing tasks (object removal, background replacement, upscaling, restoration, style transfer). It appears well thought-out from a workflow and usability perspective, but actual quality and robustness will heavily depend on how external tools (DALL-E, SD, remove.bg, Real-ESRGAN, etc.) are configured in `tools.md` and on the runtime environment. **Main strengths** - Broad functional coverage: inpainting, outpainting, background removal, upscaling, restoration, style transfer, and relighting. - Clear editing principles: non-destructive edits, working in layers, mask precision, feathering, and iterative refinement. - Concrete mappings from user intent (e.g., "remove object" or "change background") to specific techniques and tools, which should help the assistant choose an appropriate pipeline. - Good usability emphasis: the skill instructs the assistant to first clarify what edit is needed and then follow documented workflows. **Key risks and limitations** - **External dependency risk:** Actual performance and reliability depend on third-party APIs (DALL-E, SD variants, remove.bg, etc.) and correct provider setup in `tools.md`. Misconfiguration, rate limits, or model changes may degrade functionality. - **Safety and content risk:** The description does not explicitly state safeguards for sensitive or disallowed content (e.g., explicit imagery, deepfakes, harassment). If not enforced elsewhere, the skill may need additional policy-aware prompting or filtering when editing faces or generating new content. - **Quality variability:** Image realism and artifact levels will vary by tool and by quality of masks; large edits and complex composites may still require manual correction. - **Performance variability:** Latency and throughput are constrained by external services; under load or with large images, edits may be slow. **Recommended scenarios** - Assistants that already have stable connections to image-generation/editing providers and need a **general-purpose image editing orchestrator**. - Use cases that benefit from **structured, non-destructive workflows**, such as product photo cleanup, social media asset preparation, or iterative enhancement of portraits. - Users comfortable with AI-generated imagery who need **semi-automated editing** rather than pixel-perfect, professional manual retouching. **Less ideal scenarios** - Highly regulated or sensitive environments where strict, explicit image safety controls and audit trails are required. - Workflows that need deterministic, pixel-exact outputs (e.g., technical diagrams, medical imaging), where generative tools may introduce unacceptable artifacts or hallucinations.