EconTools¶
external · status: active · focus: review · discipline: economics · started: 2026
Project page: https://github.com/johanfourieza/econtools
Licence: MIT
Source: projects/landscape/econtools.yml
Positioning¶
A personal Claude Code skill pack for economic-history and economics research workflows: simulated peer review with named referee personas (Diebolt), Economic History Review journal house-style automation (EHRstyle), a research-idea sounding board modelled on a real doctoral mentor (Janluiten), a dual-model adversarial reference auditor (Kris), a LaTeX-Word round-trip for language editing (Tanniedi), and a PDF-to-markdown literature converter (Tyler). Spans ideation, literature processing, revision, referee-simulation, and dissemination.
Distinctive contribution¶
Two skills stand out versus other cataloged economics skill packs: Kris runs Claude and Codex (GPT-5.4) as two genuinely independent verification methods over a .bib file, randomly pairs references across models so no fixed Claude/Codex pair reviews the same batch, and stages an adversarial challenge round before flagging a reference as fake; EHRstyle encodes an entire journal's submission house style (Economic History Review, Feb 2026 Notes for Contributors) down to a bespoke biblatex style file. Diebolt's referee panel is explicit about its inspiration and gives public credit to two upstream projects (PHY041/claude-skill-citation-checker and Imbad0202/academic-research-skills).
Evaluation scores¶
| Dimension | Score (0–3) | Note |
|---|---|---|
| Lifecycle coverage | 2 | Five distinct stages (idea sounding board, literature conversion, dissemination formatting, citation/language revision, referee simulation) but no data-analysis, formal-modeling, or paper-drafting-from-scratch skill. |
| Autonomy level | 1 | Copilot pattern throughout: Diebolt's editor briefing requires the user to choose which changes to implement; Kris produces a scorecard for human sign-off rather than auto-editing the bibliography. |
| Architectural transparency | 3 | All skills are public Markdown with full prompt text; Kris additionally ships 8 open Python helper scripts and a labelled test fixture. |
| Inputs supported | 2 | Multiple input forms (LaTeX manuscript, .bib file, PDF folder, free-text research idea) plus literature-API access (CrossRef/OpenAlex/Semantic Scholar/arXiv and more for Kris); no private data-source access. |
| Outputs / reproducibility | 2 | Persists revised manuscripts, bibliographies, review logs (review_log.json), and a markdown literature wiki; no unified cross-skill artifact manifest. |
| Internal evaluation | 1 | Only Kris ships a labelled test fixture (10 references with known REAL/FAKE/CHIMERIC cases); the other five skills have no reported evaluation. |
| Openness | 2 | MIT license; skills are plain Markdown/Python but require per-user placeholder substitution (paths, Codex availability, Zotero setup) before they run out of the box. |
| Maturity / traction | 1 | 37 stars, 9 forks, single maintainer; active (69 commits) but pre-1.0 personal tool. |
| Cross-family policy | 1 | Diebolt and Kris both add a second model family (Codex/GPT-5.4) when available, but it is explicitly conditional ('if Codex is available') rather than required or the sole default path. |
| Runtime assurance | 2 | Kris runs two non-overlapping verification methods plus an adversarial challenge round before a reference is flagged fake; Diebolt enforces information barriers between isolated referees. |
| Cross-platform portability | 1 | Primary runtime is Claude Code, with Codex CLI as an optional second backend for two skills; no other IDEs or providers documented. |
Scored on 2026-10-01. See the evaluation rubric.
Tags¶
Pipeline stages: rq-formulation literature-synthesis revision-editing referee-simulation dissemination
Architectural features: multi-agent debate-consensus tool-use
Inputs: paper-draft bibtex-file pdf-folder research-idea
Outputs: review-report revised-bibliography formatted-manuscript markdown-literature-wiki
Knowledge sources: crossref openalex semantic-scholar arxiv orcid retraction-watch internet-archive openlibrary
Limitations¶
- Single-maintainer personal project; heavy per-user placeholder setup (paths, Codex availability, Zotero/BibTeX pipeline).
- EHRstyle is hard-coded to one journal's house style (Economic History Review) and not generalizable without rewriting the style files.
- Only one of six skills (Kris) has any labelled evaluation; the rest are undemonstrated beyond README description.
- Low external traction (37 stars) relative to the sophistication of the Kris verification design.