go-linter-driven-development v2.0.0 — rules as data

go-ldd v2 — the rules-as-data release. Each design principle now lives exactly once, as data in rules/R1–R9; skills are thin directional views over the rules, and review is done by parallel single-obsession hunters plus an over-abstraction skeptic.

⚠️ Breaking change

The quality-analyzer and go-code-reviewer agents were removed. If you invoked them directly (or reference them in your own CLAUDE.md / custom commands), migrate:

You used	Use instead
quality-analyzer	/go-ldd-analyze — read-only tests + lint + hunter/skeptic review, combined report
go-code-reviewer	@pre-commit-review — the hunter/skeptic review

Everything else is a drop-in upgrade: plugin name, all six skill names, all five existing /go-ldd-* commands, auto-detection triggers, and zero-configuration discovery are unchanged.

Upgrade

/plugin update go-linter-driven-development@ai-coding-rules

What's new

• rules/ R1–R9 — the single source of truth. Each rule ships its Principle, Why, canonical before/after, forward Design guidance, backward Fix pattern, and Falsifying questions with grep-able detection commands.
• examples/ case law — deep worked studies the rules cite: storify-leaf-type, overabstraction-cidr, dependency-rejection.
• Hunter/skeptic review — @pre-commit-review grep-prefilters the diff per rule, spawns one parallel rule-hunter per rule with hits (each fed a single rule file as its entire rulebook), then an overabstraction-skeptic that tries to kill every proposed extraction. Findings are categorized (🐛 Bugs / 🔴 Design Debt / 🟡 Readability / 🟢 Polish) and advisory — nothing blocks your commit.
• Isolated lint-fixer agent — the full-repo lint-fix loop runs out-of-conversation; mechanical issues get fixed, design failures get escalated with a rule route.
• Per-behavior TDD loop — the orchestrator now runs RED→GREEN→REFACTOR per behavior with package-scoped lint each cycle; full-repo lint happens once, in Phase 3.
• New command /wire-repo-brain [path] — bootstrap the R9 documentation network on an existing repo in one pass (code→docs edges, index.md, CLAUDE.md wiring). Run it once to make a repo AI-navigable.
• Composition-ladder testing — @testing now teaches testing each behavior at the lowest rung that contains it, with rung-tagged reusable patterns.

Full details in CHANGELOG.md and the plugin README.