China-CF-Study

What it does

Directs the agent to execute the full empirical workflow for China-context corporate finance studies, starting from a Research Proposal. It recommends 5 target journals from a China-context English-language pool, locks in a main.tex style, builds a project skeleton, and drives Python data work, LaTeX tables/figures, and template adherence. It stores decisions in a proposal file and tailors output formats to the chosen journal.

How it works

• On receiving a Research Proposal, parse and fill missing fields by prompting for the CONTRACT nine fields.
• From §3.4, generate 5 candidate journals [J1]–[J5] with topic/method/novelty/sample fit scores, plus a short justification, editorial preferences, and bib style. Output the set and stop awaiting the researcher’s selection.
• Create 00_proposal.md with top-level YAML including target_journal, bibliography_style, and writing_preferences. Build a project skeleton under the project root with a working directory structure mirroring China-CF-study guidance, including asset/, rule/, template/, and research project outputs.
• Render main.tex by replacing \bibliographystyle{} and section skeleton according to the selected journal's rules, using template/main.tex except for the header blocks.
• Begin iterative progress through sections in the order defined, comparing each completed outcome against the Expected Findings in the proposal.
• If any planned empirical result cannot be achieved, engage Plan-B logic to switch to alternative methods (e.g., different Y/X measures, FE specs, IVs, DML, PSM-DID, or robustness checks) until the project completes.
• For Plan-B, apply a predefined decision tree with failures like parallel trends, baseline coefficients, IV strength, mechanism significance, and heterogeneity, attempting prescribed substitutions and re-estimation steps.

When to use it

• Triggered only when a researcher submits a Research Proposal containing the fields listed in CONTRACT. The skill activates to drive the full lifecycle from journal targeting to deliverables and adaptive analyses.

What it can touch

• Python data analysis (pandas, numpy, statsmodels, linearmodels, econml, doubleml, scikit-learn, matplotlib)
• LaTeX outputs (main.tex, descriptive tables, baseline, measures, iv, psm_ddml, etc.) via template/ files
• The repository assets and templates located under asset/, template/, and rule/ directories are referenced to guide data sources and formatting.

Caveats

• If a proposed plan cannot meet the Expected Findings, the skill will autonomously switch to Plan-B without researcher intervention and may adjust models, data, or timelines as described in the decision tree.
• Output formats and bibliographic styling are aligned with the selected target journal, as specified by the 00_proposal.md top YAML block. All actions adhere to the 4-level structure of journal-specific sections and templates.