v0.3.0 — Plugin architecture example + Standalone→Plugin rename

A worked example, one consistent vocabulary, and a clean repo.

✨ Added

• examples/plugin-architecture/ — a worked Strict Full Plugin Architecture built inside the scaffold:
  • an enforceable system-design.md (manifest contract · load order · versioning · fault isolation · namespacing · shared-data ownership · security boundaries · contract testing · observability · collaboration patterns);
  • a runnable reference/ skeleton (Core contract + inventory/order plugins + the App composition root that assembles and runs the system);
  • two copy-paste CI gates — a manifest.schema.json and a dependency-cruiser config that blocks plugin→plugin imports.
• A new README Examples section + a Contents nav for one-page browsing.

🔁 Changed

• Renamed the "Standalone" line → "Plugin" across the kit for one consistent vocabulary: [Name]-Standalone → [Name]-Plugin, the standalone/ docs folder → plugin/, and "standalone lifecycle / standalone-first" → "plugin lifecycle / plugin-first". Lifecycle semantics are unchanged (build-independently-first → gated promotion → dependency direction main→plugin).
• Line endings normalized to LF repo-wide via .gitattributes — ends the CRLF churn and restores the docs-lint run.

---

Full changelog: https://github.com/hellOoSaksit/ai-project-scaffold/blob/v0.3.0/CHANGELOG.md · compare v0.2.1...v0.3.0

v0.2.1 — refreshed visual overview

Patch release — brings the visual overview in line with the v0.2 prompts.

Changed

• kit/principles.html refreshed to match the current prompts:
  • new Start here section — new vs refactor routing, back up before refactoring, and key/secret handling (any source)
  • rule 2 updated with any-source secret handling (never print/log/commit, never in a frontend env, rotate if leaked)
  • new Token economics section (measured: 83–96% less context per task) and anti-drift section (session-handoff · lessons · runbooks · docs-lint fails on drift · SSOT registries)

See CHANGELOG.md.

v0.2.0 — structure, security & evidence

Second release. Everything since v0.1.0, focused on structure, security, and proof.

Highlights

• OpenSSF Best Practices passing badge (project 13318) + a green docs-lint CI workflow (scripts/docs-lint.sh + Markdown link check) running on every push/PR.
• Guided intake in the scaffolder: asks new vs refactor first, then name + logic, then a tiered stack interview (recommendation + reason for each — the stack is never fixed in the prompt; it's a per-project, job-by-job choice).
• AI key/secret-handling rule in both prompts (applies to any source, not only .env) + a Security & secret-handling section in the README.
• Backup / clean-git-tree guardrail before refactoring (and a prominent backup warning in the README) — every change one undo away.
• Token economics + anti-drift README sections, backed by real, recorded evidence in docs/evidence/measurements.md (verbatim tiktoken numbers — 83–96% less context per task on a real 42-doc project — and a docs-lint pass/fail demo) plus research citations.

Changed

• Repo tidied: kit files → kit/, community-health files → .github/ (still auto-detected). Links + badge URLs updated.

See CHANGELOG.md.

v0.1.0 — AI Project Scaffold

First tagged release.

AI Project Scaffold — an AI-friendly project structure & knowledge-architecture kit for AI coding agents (Claude Code, Cursor, Copilot, Codex, Gemini CLI, Aider).

• new-project-scaffold.md — bootstrap a new project from zero
• knowledge-refactorer.md — refactor an existing project into the same shape
• principles.html — visual overview (English, offline-safe)
• MIT licensed · SECURITY / CONTRIBUTING / CODE_OF_CONDUCT · OpenSSF Scorecard CI · Dependabot · SHA-pinned actions

See CHANGELOG.md.