How hanlulong/econ-writing-skill works
1. What this repo is
- This is not an application runtime or service; it is a packaged agent skill for economics writing.
- The repository’s architecture is centered on instruction content, reference material, and installation/deployment packaging for AI assistants.
2. Primary skill payload
- The main functional unit is the econ-write skill, whose core instruction file lives in SKILL.md under both .claude/skills/econ-write/ and .agents/skills/econ-write/.
- That core file is the always-loaded “brain” of the repo: it defines writing principles, section formulas, style rules, and task modes.
3. Modular knowledge files
- The skill is split into supporting modules rather than one monolith:
- identification-strategies.md for causal inference and empirical paper types
- latex-tips.md for economics-specific typesetting and submission formatting
- review-checklist.md for self-review and audit workflows
- specialized-tasks.md for less frequent but important tasks like grants, op-eds, presentations, and referee responses
- This separation keeps the core skill lean while allowing deeper guidance when needed.
4. Platform compatibility layer
- The repo carries parallel skill directories for Claude Code and Agent Skills/Codex-style environments.
- That duplicated layout is an architectural boundary: same content, different install targets, so the skill can be discovered by multiple assistants without changing the content model.
5. Installation and packaging flow
- install.sh is the distribution mechanism that copies the skill files into the appropriate user-local skill directories.
- The README indicates support for global or project-local installation, which means the repo functions as a deployable content package rather than a build artifact generator.
6. User-facing workflows
- The skill is organized around four operational modes: drafting, rewriting, reviewing, and auditing.
- These are not separate executables; they are behavior modes encoded in the prompt/instruction hierarchy and activated by the assistant at use time.
7. Economics-writing domain coverage
- The content is structured around the major economics paper components and workflows: abstract, introduction, model, data, results, conclusion, title, and literature positioning.
- It also encodes field-specific norms for macro, trade, development, and finance, so the repository behaves like a domain knowledge system with subdomains rather than generic writing advice.
8. Empirical-method boundary
- identification-strategies.md is the main bridge between writing style and research design.
- It covers distinct empirical architectures—RCT, DiD, IV, RDD, synthetic control, event studies, structural estimation, descriptive work, and combinations thereof—so the skill can adapt its narrative structure to the method used in the paper.
9. Quality assurance and evaluation
- evals/test-cases.md and examples/before-after.md act as the repo’s validation layer.
- The examples show desired output shape, while the test cases provide a benchmark for whether the skill is producing the expected economics-writing behavior.
10. Source curation subsystem
- sources/SOURCES_RANKED.md is the provenance and authority map behind the skill.
- The README frames the skill as a synthesis of 50+ guides from prominent economists, so the repository’s knowledge base is intentionally curated and ranked rather than scraped indiscriminately.
11. Documentation and discoverability
- The README is doing architectural heavy lifting: it explains installation, supported platforms, use cases, and the internal file layout.
- In diagram terms, the README is the entry point that connects a user to the skill package, the installation script, and the supporting knowledge modules.
12. Overall system boundary
- The repo has no database, API server, or external runtime dependency chain.
- Its boundary is purely content distribution for AI assistants: source files are authored once, installed into assistant-specific directories, then consumed by the model during writing tasks.
Main entry points: econ-write/SKILL.md Core skill · econ-write/SKILL.md Core skill · README.md README · SOURCES_RANKED.md Ranked sources
Components 17
Documentation
Entry points, usage guidance, and reference material.
- README · Explains the skill, install options, supported environments, and layout.
README.md - Ranked sources · Curated authority list behind the skill's guidance and norms.
sources/SOURCES_RANKED.md
Claude Skill
Claude Code install target for the economics writing skill.
- Core skill · Primary always-loaded economics writing instruction set for Claude Code.
.claude/skills/econ-write/SKILL.md - Identification · Empirical design and paper-structure guidance for identification strategies.
.claude/skills/econ-write/identification-strategies.md - LaTeX tips · Economics-specific typesetting and submission formatting guidance.
.claude/skills/econ-write/latex-tips.md - Review checklist · Self-review and audit workflow for polishing drafts.
.claude/skills/econ-write/review-checklist.md - Special tasks · Guidance for grants, op-eds, presentations, and referee responses.
.claude/skills/econ-write/specialized-tasks.md
Agent Skill
Agent Skills/Codex-style install target for the economics writing skill.
- Core skill · Primary always-loaded economics writing instruction set for agent skills.
.agents/skills/econ-write/SKILL.md - Identification · Empirical design and paper-structure guidance for identification strategies.
.agents/skills/econ-write/identification-strategies.md - LaTeX tips · Economics-specific typesetting and submission formatting guidance.
.agents/skills/econ-write/latex-tips.md - Review checklist · Self-review and audit workflow for polishing drafts.
.agents/skills/econ-write/review-checklist.md - Special tasks · Guidance for grants, op-eds, presentations, and referee responses.
.agents/skills/econ-write/specialized-tasks.md
Validation
Examples and test cases used to check expected skill behavior.
- Test cases · Benchmarks for expected economics-writing behavior.
evals/test-cases.md - Examples · Before-and-after samples showing desired output shape.
examples/before-after.md
Packaging
Repository-level installation and distribution mechanics.
- Install script · Copies the skill into user-local assistant skill directories.
install.sh
Other components
- Writing modes · Drafting, rewriting, reviewing, and auditing behaviors encoded in the skill.
- Domain coverage · Economics paper components and subfields such as macro, trade, development, and finance.
Connections 24
- README → Install script: describes install · The README explains how the installer deploys the skill.
- README → Core skill: points to · The README introduces the main skill payload.
- README → Core skill: points to · The README explains the parallel agent-skills layout.
- README → Ranked sources: references · The README frames the curated source base behind the skill.
- Core skill → Identification: loads · Core skill is extended by the identification guidance module.
- Core skill → LaTeX tips: loads · Core skill is extended by the LaTeX guidance module.
- Core skill → Review checklist: loads · Core skill is extended by the review workflow module.
- Core skill → Special tasks: loads · Core skill is extended by the specialized tasks module.
- Core skill → Identification: loads · Core skill is extended by the identification guidance module.
- Core skill → LaTeX tips: loads · Core skill is extended by the LaTeX guidance module.
- Core skill → Review checklist: loads · Core skill is extended by the review workflow module.
- Core skill → Special tasks: loads · Core skill is extended by the specialized tasks module.
- Core skill → Core skill: mirrors · The two core skill directories carry the same content for different install targets.
- Identification → Writing modes: shapes · Identification guidance informs drafting and rewriting behavior.
- Review checklist → Writing modes: governs · Review guidance governs audit and polishing workflows.
- Special tasks → Domain coverage: covers · Specialized task guidance expands domain and output coverage.
- LaTeX tips → Domain coverage: formats · Formatting guidance is tied to economics submission conventions.
- Ranked sources → Core skill: informs · Ranked sources provide the authority base for the core skill.
- Ranked sources → Core skill: informs · Ranked sources provide the authority base for the mirrored core skill.
- Examples → Test cases: complements · Examples and test cases together define expected output behavior.
- Test cases → Core skill: validates · Test cases check whether the core skill behaves as intended.
- Examples → Core skill: illustrates · Examples show the target shape of the skill's outputs.
- Install script → Core skill: deploys · Installer copies the Claude-targeted skill package.
- Install script → Core skill: deploys · Installer copies the agent-skills-targeted package.