inno-prepare-resources

The inno-prepare-resources skill acts as a foundational component in the Oh My Paper research pipeline, designed to bridge the gap between abstract research intent and concrete project implementation. It is primarily intended for researchers and AI agents using Claude Code to automate the initialization phase of a research project. By programmatically reading an evaluation instance, this skill orchestrates the retrieval of necessary reference material, ensuring that subsequent experimental stages are grounded in reliable upstream evidence and existing scholarly work.

• Automatically processes evaluation instances to extract paper lists, task requirements, and metadata.

• Performs targeted GitHub searches to identify relevant reference repositories, aiding in baseline establishment.

• Generates structured dataset descriptions and imports metaprompt modules for custom or pre-existing datasets.

• Automates the bulk download of arXiv paper sources based on titles, populating local project reference directories.

• Maintains traceability by persisting intermediate results, such as GitHub search logs and agent output, into specified log subdirectories.

• Integrates with local research project structures, utilizing absolute path resolution to ensure consistency across different environment setups.

• Ensure the instance.json file is correctly populated before execution to avoid initialization errors.

• The skill supports two primary operating modes: Idea mode and Plan mode; provide the corresponding 'ideas' input in Plan mode for more precise repository selection.

• All generated artifacts are saved to the active project workspace rather than the skill directory, maintaining clear separation between logic and data.

• If required APIs or dependencies are unavailable, the skill provides descriptive feedback, allowing for manual fallback instead of silent failure.

• Operates under a strict contract of inspection before mutation, preserving existing research structures and metadata.