Sync main branch: v3.1 → v4.1.1 + UAI project setup - #2
Open
thecopybookhare-cmd wants to merge 48 commits into
Open
thecopybookhare-cmd wants to merge 48 commits into
thecopybookhare-cmd wants to merge 48 commits into
Conversation
No embedded titles in ggplot or generated .tex files — use descriptive
filenames and LaTeX \caption{}. Enforces booktabs, threeparttable, serif
fonts, publication-quality labels, and full year axes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add figures.md and tables.md to rules table in customization guide. Update version badge in README. Add changelog entry for 2.0.2. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add figures and tables rules (v2.0.2)
Enforces standard economics working paper formatting: - 12pt article, 1in margins, double-spaced body - Title page: titling package for proper font sizes, \quad authors, single-spaced abstract - Required packages, table/figure conventions, bibliography format - Writer-critic deduction rubric for format violations
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
v2.0.3: Working paper format rule
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ools format Two bugs in all 10 skill SKILL.md files: 1. `disable-model-invocation: true` was set on every skill, preventing Claude from auto-dispatching skills when the user describes a task. This defeats the core UX — users should be able to say "review my paper" and have Claude pick `/review` automatically. 2. `allowed-tools` used JSON array syntax (`["Read", "Grep"]`) instead of the comma-separated string format (`Read,Grep`) that Claude Code expects. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Fix skill frontmatter: auto-invocation and allowed-tools format
…eview simulation Scope expansion from economics to all empirical social science (finance, marketing, management, accounting, public policy). Complete redesign of peer review simulation with editor agent, referee dispositions, pet peeves, R&R memory, and desk reject. 71% reduction in per-session context overhead. Key changes: - New editor agent with desk review, literature verification (WebSearch), referee selection based on journal culture, FATAL/ADDRESSABLE/TASTE classification, and R&R memory across rounds (max 3) - 6 referee dispositions weighted by journal culture + balanced pet peeves - 35 journal profiles across 10 field sections (economics, finance, accounting, marketing, management) - Folder restructure: paper/ is self-contained (figures, tables, talks, quarto, preambles, supplementary, replication all inside) - Quarto RevealJS presentation support alongside Beamer - Table type templates (descriptive stats, regression, multi-outcome panels, balance, robustness) - Figure standards (colorblind palettes, grayscale independence, PDF output) - Agent prompts made field-neutral (read from domain-profile, default economics) - Auto-memory replaces [LEARN] tags - Archives deleted (22 rules, 16 agents, 20+ skills, 4 hooks) - Guide and documentation fully updated for v3.0 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add 3.0 branding to guide homepage. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- /tools upgrade: automated infrastructure upgrade preserving user content - Migration guide added to README and user guide - Updated /tools deploy for gh-pages workflow Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…files are safe Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The v2.0→v3.0 consolidation accidentally stripped practical detail from 6 skills. This restores all 22 lost items: proximity scoring, feasibility grades, PAP templates, 12-category code checklist, 4-phase econometrics protocol, replication tolerances, context checklists, and more. Also expands referee pet peeves from 24 to 51 (27 critical + 24 constructive) for more realistic and varied peer review simulations. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All references to Paper/, Figures/, Tables/, Talks/, Preambles/, Replication/, Supplementary/ updated to lowercase paths inside paper/ (paper/figures/, paper/tables/, paper/talks/, paper/replication/, etc.) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Git was still tracking uppercase Paper/ despite the filesystem rename. Force re-index to lowercase paper/ to match all path references. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
/new-project now enters plan mode before starting Discovery. Creates folder structure, drafts plan, waits for approval. Also fixed double-nested path .claude/rules/.claude/references/. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
CLAUDE.md now has an Output Organization field (by-script or by-purpose). Default is by-script: outputs go to subfolders named after the generating script (e.g., paper/figures/main_regression/figure1.pdf). Coder agent reads this setting and follows accordingly. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace natbib/bibtex with biblatex/biber (natbib=true compat preserved) - Add microtype for improved typography - Add fancyhdr for clean page numbering - Add colored citations (azure RGB 0,127,255) - Add captionsetup (small font, bold labels) - Add custom column types (L, C, R) - Add math operators (argmax, argmin, norm, indicator) - Add URL bleeding fixes for bibliography - Require hyperref loaded last - Full reference preamble included as copy-paste standard Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The \DeclareUnicodeCharacter command is for pdflatex/inputenc only.
Under xelatex it renders as literal text ("22C6★") on a phantom page.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…iblatex Update working paper format: biblatex + polished preamble
Adversarial review flagged the original RTF file with two false claims (Bibliography_base.bib removed, Referee pool missing) and one already-fixed issue (Talks/ stale paths). Rewritten as clean Markdown reflecting reality. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…case paths - Delete quality_score.py (761 lines targeting nonexistent Quarto/Slides/) - Delete sync_to_docs.sh (92 lines targeting nonexistent deployment structure) - Remove sync_to_docs.sh permissions from settings.json, add biber permission - CLAUDE.md: bibtex → biber in compilation commands - tools/SKILL.md: bibtex → biber, Paper → paper, Talks → paper/talks, Preambles → preambles Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ics, README honesty Tables: - Significance stars now journal-dependent (AEA: no stars; others: default) - Journal profiles gain Table format field for overrides - content-standards.md drops false "AER, QJE, Econometrica" universal claim Format: - working-paper-format.md splits Required (blocking) from Recommended (advisory) - Writer-critic deductions adjusted for recommended items Scope: - "Empirical social science" → "Economics" across README, CLAUDE.md, guide, meta-governance - Other fields can adapt via domain profile and journal profiles README: - Strip pitch language, add Limitations section - "Simulated" not "Realistic" for peer review - "Configured" not "calibrated" for referee pools Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Writer (101→307 lines): paragraph-level argument moves, 4 paper-type section templates, design-specific guidance, humanizer demoted to cleanup. Writer-critic (111→208 lines): 8 check categories, paper-type coherence, design-specific completeness tables, results narration checks. Strategist (70→264 lines): structural estimation strategy, theory+empirics predictions, descriptive measurement strategy, design-specific estimation. Strategist-critic (302→457 lines): structural model checklists, theory prediction sharpness, descriptive construct validity, type-specific sanity. Coder (102→355 lines): paper-to-code naming maps, numerical discipline (float guards, CDF clamping, integer literals, pre-allocation), structural implementation (MLE/GMM/SMM/BLP), Monte Carlo patterns, prohibited patterns. Coder-critic (160→278 lines): 16 check categories, numerical discipline enforcement, structural code checks, simulation code checks. Methods-referee (169→207 lines): paper-type-specific evaluation dimensions and scoring rubrics for reduced-form, structural, theory+empirics, descriptive. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…, path fixes agents.qmd: Updated descriptions for all 7 rewritten agents — paper-type awareness, argument moves, numerical discipline, check category counts. architecture.qmd: Updated execution phase tables to reflect new agent capabilities (16 coder-critic checks, 8 writer-critic checks). user-guide.qmd: Rewrote /write section to describe paragraph-level argument moves and paper-type-specific templates. Updated /review descriptions. reference.qmd: Updated /write with new section options (data, model), /strategize with paper-type awareness. Fixed output path descriptions. All guide files: Fixed stale Paper/ → paper/ paths across all .qmd files. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Each agent group gets a callout box with candid assessment of what works and what doesn't: - Discovery: Librarian misses non-indexed papers; Explorer can't verify access - Strategy: Good at checklists, weak at institutional judgment; structural checklists are newer and less tested - Execution: Writer produces drafts not prose; Coder can't debug runtime; critic checks structure not substance - Review: Catches structural problems, can't replicate field expertise; desk reject depends on web search quality - Infrastructure: Scores are deduction rubrics not quality measures - Talks: Optimizes for completeness, the opposite of good slide design Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New reference files derived from C++ Core Guidelines engineering discipline: - coding-standards-r.md: data.table idioms, renv, future.apply, prohibited patterns - coding-standards-python.md: type hints, NumPy discipline, black/ruff, joblib - coding-standards-julia.md: type system, multiple dispatch, in-place operations Each covers: numerical discipline (float guards, CDF clamping, pre-allocation), function design (fail fast, consistent API, documentation), parallelism patterns, prohibited patterns with reasons and replacements. Coder agent now reads the language-specific standards before writing code. Removed stale journal-profiles.html (nothing referenced it). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
changelog.qmd surfaces v4.0.0 release notes (paper-type architecture, numerical discipline, scope narrowing, table standards, dead code removal) plus summaries of v3.x releases. Added to navbar. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Storyteller: paper-type-specific narrative arcs (structural talks present counterfactuals as payoff, theory talks lead with distinguishing prediction, descriptive talks lead with data innovation). Storyteller-critic: 6 check categories (was 5), adds paper-type coherence check and type-specific arc validation with deduction table. Guide: front page title 3.0 → 4.0. Callout boxes rewritten from self-deprecating paragraphs to brief, balanced notes. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Storyteller (86→338 lines): paper-type narrative arcs, full Beamer design guide (progressive reveal, columns, highlighting, table simplification), full Quarto RevealJS guide (auto-animate, tabsets, custom SCSS, speaker notes, scrollable backups). 12 slide design principles including visual rhythm, 3-second test, and "design for the room." Storyteller-critic: 6 checks with paper-type coherence. Guide: front page 4.0, callout boxes brief and balanced. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add mechanical enforcement underneath the existing agent architecture.
Agents still do judgment work; the new layer catches grep-able violations
before judgment is even needed.
Five changes:
1. Content invariants (.claude/rules/content-invariants.md)
- 21 numbered non-negotiable rules across Paper (INV-1–13),
Code (INV-14–19), and Talk (INV-20–21)
- Consolidates scattered rules from content-standards, working-paper-format,
and coding standards into one canonical reference
- writer-critic, coder-critic, storyteller-critic, verifier all updated
to cite invariant numbers in reports
2. Pre-flight reports (strategize + analyze)
- Strategist must output Pre-Strategy Report proving it read literature
review, data assessment, and research spec before designing anything
- Coder must output Pre-Code Report proving it read strategy memo and
established paper-to-code naming map before writing any code
- Both agents and skills updated with mandatory first output
3. Grep-based linter (/tools lint + lint-scripts.sh)
- 20 mechanical checks across R (12), Python (6), Julia (3) + shared
- Draws from coding-standards-*.md prohibited patterns
- /review --code now runs lint as Step 1 before dispatching coder-critic
- Portable (macOS grep -E, no Perl regex)
4. Decision records (discover + strategize)
- ADR-style records at Discovery (research question choice) and
Strategy (identification method choice)
- Captures what was chosen, what was rejected and why, key assumptions,
and what would invalidate the decision
- Shared template: templates/decision-record.md
- Saves to quality_reports/decisions/
5. PostToolUse lint hook
- Auto-runs linter on every Edit/Write to R/Python/Julia files
- Advisory (exit 0), skips .claude/ internals
- Wired into .claude/settings.json
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two passes on the guide website:
1. Content sync to v4.1.0:
- Changelog: added v4.1.0 entry
- Reference: added /tools lint and /tools upgrade
- Customization: 5 hooks (was 4), fixed rules list, added content-invariants
- User guide: added pre-flights, decision records, lint mentions
- SCSS: prefers-reduced-motion, reduced scanline opacity
2. UX restructure (scored 6/10 in usability audit):
- user-guide.qmd rewritten from 370-line feature dump to 210-line
walkthrough with fabricated output examples (editorial decision,
argument-move structure)
- index.qmd: "Just want 2 commands?" callout, trimmed jargon,
collapsible Day 1 checklist
- agents.qmd: nested paragraphs collapsed into paper-type tables,
coder-critic score breakdown example added
- architecture.qmd: removed duplicated config table and stale sections
(slides/lectures, constitutional governance)
- customization.qmd: received exploration workflow, self-improvement,
and constitutional governance sections from other pages
- De-duplicated: each concept now has one canonical page, links elsewhere
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Modernize LaTeX infrastructure with Overleaf-compatible upgrades: - latexmk build system replaces manual 4-command compilation - tabularray for hand-written tables (R output keeps bare tabular) - cleveref for smart cross-references (\cref instead of Figure~\ref) - microtype promoted from Recommended to Required - GitHub Actions CI for paper compilation on push/PR Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- CI: remove extra_system_packages (fontspec is a LaTeX pkg, not Alpine); add main.tex existence check for template repo graceful skip - Guide: add v4.1.1 changelog, update reference/user-guide/customization for latexmk, tabularray, cleveref, microtype promotion - Changelog: update v4.0.0 Required list to reflect current state Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Set Institution to Universidad Adolfo Ibañez and Field to Business Administration for this project instance. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- CHANGELOG.md: keep v4.1.0/v2.0.2 link definitions from main - docs/*.html + search.json: keep deletions from main (v4.x reorganization) - Username already correct (hugosantanna) in main Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Transfers complete research from claude/silly-satoshi worktree to main so files are accessible in any session without worktree context. Includes: MASTER_DATA.md, RESUME_HERE.md, all Factiva findings, NotebookLM audits, lit review, data exploration, legal gaps, professor feedback, research spec, and addenda. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
mainbranch (v3.1 → v4.1.1) to the remote — 46 commits spanning major version upgradesWhat's included
latexmk,tabularray,cleveref, CI fixTest plan
🤖 Generated with Claude Code