SIMON / PERSONAL · MIT

Local image repair

Reconstruct a selected region of an owned or authorized image using IOPaint and LaMa.

Workflow and wrapper by Simon Lin; underlying tools and models belong to their upstream projects.

01 / REQUIREMENTS

Before you start

Python 3.11, Pillow, IOPaint and separately downloaded LaMa model.

  • Explicit rectangular or corner region
  • Originals preserved; PNG output
  • Duplicate-name and missing-output checks

02 / INSTALL

Install and use

unzip image-repair-v1.1.0.zip -d ./skill-install
mkdir -p "$HOME/.agents/skills/image-repair"
cp -R ./skill-install/image-repair/. "$HOME/.agents/skills/image-repair/"

For Claude Code, use .claude/skills instead of .agents/skills. For other clients, follow their loading convention. Keep commands consistent with your actual install path, then read SKILL.md and THIRD-PARTY.md.

Use repair.sh with an explicit region and separate output directory, following SKILL.md. Keep originals and inspect repaired regions.

No dependencies, model weights, accounts, authentication or media are bundled. Upstream services may have their own costs and usage limits.

03 / UPSTREAM & LICENSE

Attribution and checks

Original documents and wrappers: MIT, copyright PhoenixTree Holdings Limited; author Simon Lin. Upstream licenses and model terms remain separate.

Original package snapshot: 2026-08-18. Free distribution: 2026-10-22. Documentation, script syntax, file inventory and archive integrity checked on 2026-10-22 (Asia/Singapore). Full model/browser workflows and all-client compatibility were not rerun.

Package files

  • LICENSE.txt1115 B
  • README.md2787 B
  • SKILL.md2113 B
  • THIRD-PARTY.md250 B
  • repair.py4931 B
  • repair.sh363 B