Test repository for action-style-guide — contains test lectures and development scripts for validating the QuantEcon Style Checker.

This repository provides:

1. Test lectures with intentional style violations for regression testing
2. Clean lectures that should pass all checks (baseline validation)
3. Development scripts for prompt iteration and ground truth validation
4. GitHub workflows to test the action in a production-like environment

The primary test lecture. Contains 42 carefully embedded violations:

• Writing (12): Multi-sentence paragraphs, wordiness, capitalization, title case, bold/italic, visual elements
• Math (5): Transpose notation, matrix brackets, equation formatting, unicode parameters
• Code (6): Package installation, Greek letter unicode, timing patterns
• JAX (5): Functional programming, NamedTuple, PRNG key management
• Figures (5): Line width, titles, captions, spines
• Links (3): Link text quality, URL formatting
• References (2): Citation syntax
• Admonitions (4): Exercise formatting, nesting, directive syntax

See lectures/markov_chains_jax.annotated.md for the complete violation catalog.

Standalone script for prompt iteration testing. Tests a single rule + prompt against a lecture file without needing the full qestyle installation.

# Requires ANTHROPIC_API_KEY environment variable
python scripts/test_rule_prompt.py lectures/markov_chains_jax.md

Features:

• Edit PROMPT and RULE variables directly in the script to iterate
• Extended thinking toggle (USE_EXTENDED_THINKING), configurable budget
• Counts violations and identifies false positives (identical current/suggested text)
• Saves raw LLM response to scripts/last_response.md for comparison

Deterministic ground truth finder for qe-writing-001 (one sentence per paragraph) violations.

python scripts/find_multisentence.py lectures/markov_chains_jax.md

Parses the file structurally (skips code blocks, directives, frontmatter) and reports all paragraph blocks containing multiple sentences. Useful for validating LLM recall against a known baseline.

Install qestyle from the main repo, then run against test lectures:

# Install qestyle
pip install git+https://github.com/QuantEcon/action-style-guide.git

# Test writing rules
qestyle lectures/markov_chains_jax.md --categories writing

# Dry run (report only, no changes)
qestyle lectures/markov_chains_jax.md --dry-run --categories math

# Reset test files after testing
git checkout -- lectures/

Create an issue with a comment like:

@qe-style-checker lectures/test-lecture-violations.md math,code

The action can be triggered from the Actions tab to test specific categories.

• action-style-guide — The main style checker action
• docs/testing-extended-thinking.md — Extended thinking experiment results