@jaimevalasek/aioson 1.7.2 → 1.9.0

package/CHANGELOG.md

@@ -1,560 +1,595 @@
-# Changelog
-
-All notable changes to this project will be documented in this file.
-
-## [Unreleased]
-
-## [1.7.2] - 2026-04-07
-### Fixed
-- Removed 5 spurious Windows alternate data stream files (`SKILL.md:Zone.Identifier`) from `template/.aioson/skills/design-system/` that leaked from a local Windows copy operation and were never tracked by git.
-- `tests/agent-contracts.test.js`: added `neo`, `sheldon`, `tester`, and `orache` to the full locale coverage check; added a separate `BASE_ONLY_AGENTS` list (`copywriter`, `design-hybrid-forge`, `site-forge`) with a dedicated test that verifies their base files ship in the template. Previously these agents were not validated by the test suite.
-
-## [1.7.1] - 2026-04-07
-### Added
-- **Marketing & copywriting system**: new `@copywriter` agent (5 operating modes including VSL scripts) backed by a generic `copywriting` genome (One Belief, 5 Acts, PMS framework, market sophistication levels, 10 heuristics).
-- **Marketing references library** (`.aioson/skills/marketing/references/`): 8 on-demand reference files with 70+ patterns and anti-patterns covering One Belief, 5-Act narrative structure, fascinations (12 formulas), offer structure, PMS research, copy patterns, anti-patterns checklist, and market intelligence tools. Loaded conditionally by `@copywriter` to keep context lean.
-- **VSL Craft skill** (`.aioson/skills/marketing/vsl-craft.md`): Video Sales Letter production playbook — 3 formats (horizontal/vertical/hybrid), 5-act script structure, 5 hook formulas, retention techniques, testing methodology, and production specs.
-- **Marketing integration across agents**: `@neo` routes to `@copywriter` when `project_type=site` and copy is missing; `@setup` registers the copywriting genome and adds copy gate to site routing; `@dev` enforces a copy guard for marketing pages and reads the 5-Act structure from `copy-{slug}.md`; `@ux-ui` maps the 5 Acts to visual sections; `@deyvin` routes marketing copy requests to `@copywriter` instead of writing inline.
-- **Landing page production skills**: `landing-page-forge.md` (animation libraries, performance, SEO/LLMO, tracking, 3-track parallel production checklist) and `landing-page-deploy.md` (Vercel and Hostinger VPS deploy paths).
-- **Digital marketing agency squad template** under `template/.aioson/templates/squads/digital-marketing-agency/`.
-
-### Changed
-- `@dev` site production section now includes a copy guard that blocks inline marketing copy on `project_type=site` pages — copy must come from `@copywriter` via `copy-{slug}.md`.
-- `@ux-ui` copy gate now reads the 5-Act narrative structure from copy files and maps acts to UI sections.
-- `@neo` workflow stage detection includes `Needs copy` for site projects without a copy file.
-- `@setup` routing table splits `site` into two paths: `@copywriter → @ux-ui → @dev` (no copy yet) vs `@ux-ui → @dev` (copy ready).
-- Misc updates to `@qa`, `@squad`, `squad-create`, `squad-design`, `cognitive-core-ui/motion`, `ui-ux-modern`, and `template/CLAUDE.md`.
-
-### Notes
-- Marketing references are loaded conditionally by `@copywriter` only — other agents never load them directly. This keeps the marketing system isolated and prevents contaminating non-site projects.
-- The copy guard triggers exclusively for `project_type=site` marketing/sales pages. SaaS UI, dashboards, app interfaces, and documentation sites use the standard implementation flow with no behavior change.
-
-## [1.7.0] - 2026-04-06
-### Added
-- **Squad autonomous execution system**: full multi-agent squad stack — intra-squad message bus (`intra-bus.js`, `squad:bus` command), executor reflection module (`reflection.js`), task decomposer, `squad:autorun` for end-to-end autonomous execution, learning system, squad daemon, verify-gate, and cross-AI synthesis. Intelligence gaps addressed in plans 80–82; 6 critical operational fixes in plan 83.
-- **Runner system**: `runner:run`, `runner:queue`, `runner:plan`, and `runner:daemon` commands for persistent background job execution outside the main session loop.
-- **New CLI commands**: `agent:audit` (agent integrity inspection), `brief:gen` (context brief generation), and `verify:gate` (execution pre-flight gate). Documented in `docs/pt/`.
-- **SDD automation — The 80% Rule (plans 74–79)**: automation scripts that drive spec compliance across agents; harness sensors, context budget ceiling, and PGE pattern (plan 76); full SDD coverage across all agents (plan 77); locale SDD sync and pt-BR `tester.md` (plan 78); SDD automation scripts wired into agents and docs (plan 79); engine fixes for `evaluateReadiness`, `extractLastCheckpoint`, and `detectTestRunner`.
-- **Agent capability sprints**:
-  - Sprint 1 — task-list working memory, hook contract, file size guidelines, `CLAUDE.local.md` support.
-  - Sprint 2 — context compaction protocol and self-directed plan mode.
-  - Sprint 3 — CronTools protocol and config tiers documentation.
-- **Web-research-cache skill**: caches web research results in `researchs/`; awareness propagated to all agents that perform web discovery.
-- **`site-forge` agent**: replaces `hybrid-clone` with a dedicated site-forge agent backed by a Brains knowledge system for site cloning and forging workflows.
-- **Three.js skill**: `threejs-spatial` modifier and corresponding agent (`hybrid-clone`) for Three.js spatial interface projects.
-- **Operational pipeline**: event enrichment, spec-sync, token economy, devlog pipeline, and project hooks system.
-- **Squad CLI integration**: `squad.md` wired to CLI tools — Step 0 scaffold, CLI integration table.
-- **Tutorials site**: squads tutorial page added; Squads and Automation links activated on the tutorials index.
-
-### Changed
-- `update` now only updates already-installed files by default; use `--all` to sync every template file.
-- Setup routing no longer directs to `@dev` when only `plans/` or research files are present.
-- `@deyvin`, `@neo`, and `@sheldon` now load and maintain `dev-state.md` awareness.
-- README redesigned with visual impact, squads ecosystem section, and spec-driven workflow documentation.
-
-### Fixed
-- `@orache` research cache: agent now writes to `researchs/` (previously only read from it).
-- SDD engine reliability: `evaluateReadiness`, `extractLastCheckpoint`, `detectTestRunner` (plan 79 phase 2).
-- Squad autonomous operation: 6 critical fixes for stable real-world execution (plan 83).
-- Three.js patterns: race condition and invalid property reference resolved.
-
-## [1.6.0] - 2026-04-01
-### Added
-- **Interactive install wizard**: animated multi-screen onboarding (`init` and `install`) with Design (screen 3) and Locale (screen 4) dimensions, multi-select for design skills, and pre-populated choices on `--reconfigure`.
-- **`aioson setup` command**: unified setup entry point with auto-detected system language.
-- **Spec-driven process skill** (`aioson-spec-driven`): agent contract upgrades, GSD lessons integration (must_haves, 4-tier verification, gray areas, seeds, forensics), and context budget warning threshold configuration.
-- **Design skills**: `glassmorphism-ui` and `neo-brutalist-ui` added; HTML preview gallery for all 8 design skills with per-skill landing page previews.
-- **Aurora Command UI hybrid**: design hybrid variation workflow and aurora-command-ui skill.
-- **5-phase context optimization system**: implemented in `src/commands/context-optimizations.js`.
-- **Health digest, dynamic tools, and learning evolve pipeline**: `--all` flag for the main update flow; health digest output and adaptive learning pipeline wired to the operational surface.
-
-### Changed
-- `install --reconfigure`: warns when the new profile deselects previously installed items; confirm screen strings fully localized.
-- `update`: now preserves locale from the saved profile and installs all framework files when upgrading; added `--all` flag.
-
-### Fixed
-- `install`: `overwrite=true` enforced when reconfigure changes the active profile.
-- `design-hybrid` options: TTY error logged correctly; test coverage expanded.
-- `install-wizard`: broken banner box, wrong skip reason, and stdin hang after wizard completion resolved; terminal state now fully restored.
-- Landing page previews: hero content and section titles centered correctly across all 3 previews.
-
-## [1.5.1] - 2026-03-28
-### Added
-- **TDD Gate no `@dev`**: novo bloqueador explícito antes de qualquer implementação de lógica de negócio. Detecta test runner via `test_runner` em `project.context.md` ou varre a raiz em busca de `pest.xml`, `vitest.config.*`, `pytest.ini`, `.rspec`, `foundry.toml`. Mandato RED → GREEN → commit por classificação (MICRO/SMALL/MEDIUM) com exceções para migrations, configurações e conteúdo estático.
-- **Novo agente `@tester`**: engenheiro de testes para aplicações já implementadas. Produz `test-inventory.md` (mapa de cobertura), `test-plan.md` (estratégia escolhida + cobertura antes/depois) e escreve testes por módulo em ordem de risco. Suporta Characterization Testing, Test Pyramid, Risk-first Gap Filling, Property-based, Contract e Mutation Testing. Mapeamento completo de frameworks: Laravel/Pest, Next.js/Vitest+RTL, Django/pytest, FastAPI/httpx, Rails/RSpec, Solidity/Foundry.
-- **Detecção de documentos de kickoff no `@product` e `@sheldon`**: ambos os agentes agora varrem `plans/*.md` e `prds/*.md` na raiz do projeto antes de iniciar o fluxo. Detectam contexto (greenfield vs feature), oferecem sintetizar o conteúdo nos artefatos formais em `.aioson/context/`, e nunca modificam os arquivos fonte — o usuário decide quando deletá-los.
-- Campo `test_runner` adicionado ao Context contract em `config.md` e ao template de `project.context.md` gerado pelo `@setup` (todos os 4 locales).
-- Detecção automática de test runner no `@setup`: varre a raiz e preenche `test_runner` em `project.context.md` quando um runner é detectado.
-- `@tester` registrado em `CLAUDE.md`, `AGENTS.md`, `OPENCODE.md`, `.gemini/commands/aios-tester.toml`, `src/constants.js` (AGENT_DEFINITIONS) e `template/.claude/commands/aioson/agent/tester.md`.
-- Documentação `docs/pt` atualizada: `agentes.md` (seções `@dev`, `@qa`, `@product`, `@sheldon` novo, `@tester` novo, fluxos atualizados), `cenarios.md` (exemplo TDD no Cenário 2, exemplo kickoff doc, nota @tester vs @qa), `inicio-rapido.md` (tabela de sequências e seção de estratégia de kickoff docs).
-
-### Changed
-- `src/context-writer.js` e `src/commands/setup-context.js`: suporte ao campo `testRunner` na geração e override de `project.context.md`.
-- `template/.aioson/agents/dev.md`: seção "Implementation strategy" reescrita com mandato RED→GREEN→commit; seção "Atomic execution" atualizada com referência cruzada ao TDD Gate.
-- `template/.aioson/agents/product.md`: seção "Source document detection" adicionada antes de "Mode detection".
-- `template/.aioson/agents/sheldon.md`: seção "Detecção de documentos fonte" adicionada antes de RF-01.
-- `template/.aioson/config.md`: campo `test_runner` no Context contract; `@tester` no workflow opcional.
-- Versão de referência nos templates de `project.context.md` atualizada de `0.1.25` para `1.5.1`.
-
-## [1.5.0] - 2026-03-27
-### Added
-- Native web map and scrape commands for first-party web discovery and extraction workflows.
-- New agent capabilities and orchestration surface including `@neo`, the Sheldon PRD enrichment flow, richer downstream gates, and improved locale-aware runtime execution rules.
-- Squad-level operational improvements including webhook/channel bridge integration, daemon and registry evolution, and dashboard/design-system refinements for the squad experience.
-- Cognitive Core UI design skill upgrades with stronger art direction, composition libraries, anti-generic guardrails, and more robust interaction/contrast guidance.
-
-### Changed
-- Project license changed from `MIT` to `AGPL-3.0-only` for new releases going forward.
-- Package metadata and lockfile versioning are now aligned at `1.5.0`.
-- Agent workflows now enforce PRD gating, stronger atomic execution behavior, and more explicit routing for MICRO and product/design-driven flows.
-
-## [1.3.0] - 2026-03-15
-### Added
-- Canonical runtime observability for installed projects: runtime SQLite now acts as the primary execution store with task/run/event lifecycle tracking, workflow event synchronization, and an execution gateway prepared to instrument direct agent entrypoints, workflow enforcement, squads, and operational runs from the project runtime itself.
-- `docs/pt/comandos-cli.md`: first Portuguese-first CLI reference page covering all public `aioson` commands with grouped descriptions, command map, examples, and practical usage recipes. Linked from `docs/pt/README.md` and `docs/pt/inicio-rapido.md`.
-
-### Changed
-- Full product rebrand from `aioson` to `aioson` across package metadata, CLI branding, docs, templates, prompts, tests, runtime paths, and generated project structure. The managed project directory is now `.aioson/` instead of `.aioson/`.
-- npm distribution moved to the scoped package `@jaimevalasek/aioson` after the unscoped `aioson` name was rejected by npm for similarity policy reasons. Install and `npx` docs now use the scoped package while the installed CLI commands remain `aioson` and `aios`.
-- Release description updated to `AI operating framework for hyper-personalized software.` to match the current platform direction.
-- Legacy `dashboard:*` CLI flow removed from the public surface. `aioson dashboard:init|dev|open` no longer appear in help or docs; if called, the CLI now returns a migration error instructing users to open the separately installed dashboard app and select the project folder that already contains `.aioson/`. Updated pt docs and agent guidance to match the new app-first dashboard flow.
-- `template/.aioson/skills/static/interface-design.md`: upgraded the interface-design craft guide by merging strengths from the upstream `interface-design` skill without losing AIOSON workflow fit. Added persistent design memory via `.interface-design/system.md`, existing-UI refinement rules (diagnose drift before re-theme), two fully specified directions (`Sophistication & Trust`, `Premium Dark Platform`), a mandatory pre-component decision checkpoint, and explicit memory update guidance for reusable UI patterns.
-
-## [0.1.28] - 2026-03-04
-### Added
-- **`scan:project` command** (`aioson scan:project [path] [--provider=<name>] [--dry-run] [--json]`): Node.js port of the brownfield project scanner, replacing `aioson-scan.py`. Walks the project, reads key files, calls a cheap LLM (multi-provider), and generates `discovery.md` + `skeleton-system.md`. Zero npm dependencies — uses `node:fs`, `node:https`, `node:http` only.
-- Multi-provider HTTP client supporting OpenAI-compatible format (DeepSeek, OpenAI, Gemini, Groq, Together, Mistral) and Anthropic native format. Configured via `aioson-models.json` (same config file as before).
-- `--provider=<name>` flag to override `preferred_scan_provider` from config at runtime.
-- `--dry-run` mode: walks the project and reports file counts without calling the LLM.
-- i18n keys for `scan_project` section in all 4 locales (en, pt-BR, es, fr).
-- `install` hint message updated in all 4 locales: `python aioson-scan.py` → `aioson scan:project`.
-
-### Removed
-- `template/aioson-scan.py`: Python scanner removed from the template. `aioson scan:project` is the normalized replacement. The `aioson-models.json` config file (provider API keys) is unchanged.
-
-### Changed
-- All 15 template agent files (analyst, dev, orchestrator — base + 4 locales each): `python aioson-scan.py` replaced with `aioson scan:project` in brownfield detection hints.
-
-## [0.1.27] - 2026-03-04
-### Added
-- **HTML reports** for browser QA: `--html` flag added to `qa:run`, `qa:scan`, and `qa:report`. Generates a self-contained visual report in `reports/<date>_<time>_<mode>/index.html`. Screenshots embedded as base64 — fully portable for sharing. Existing `aios-qa-report.md` and `aios-qa-report.json` outputs are never modified; HTML is purely additive.
-- `reports/index.html`: historical index auto-updated after each HTML run, listing all runs with date, mode, URL, and severity badges. Sorted newest first.
-- `qa:report --html`: retroactively generates HTML from the existing `aios-qa-report.json` without re-running Playwright.
-- HTML report features: severity filter buttons, collapsible finding cards (location + risk + fix + screenshot), performance cards with colour-coded thresholds, AC coverage table, routes list (scan mode), click-to-zoom screenshot lightbox, vanilla JS (no framework).
-- `src/qa-html-report.js`: self-contained HTML generator module with embedded CSS and JS, `writeHtmlReport()` and `updateReportsIndex()` exports.
-- i18n key `html_report_written` added to `qa_run`, `qa_scan`, `qa_report` sections in all 4 locales (en, pt-BR, es, fr).
-- Missing `help_qa_*` i18n keys added to pt-BR locale (were absent; es/fr fall back to en).
-- `docs/en/qa-browser.md`: `--html` flag documented for all three commands, plus a new **HTML reports** section describing folder structure, features, and `qa:report --html` retroactive mode.
-- README: `--html` flag added to Browser QA command signatures and Agent usage helper examples.
-
-## [0.1.26] - 2026-03-04
-### Added
-- **Browser QA engine** (`qa:init`, `qa:doctor`, `qa:run`, `qa:scan`, `qa:report`): full Playwright-powered browser testing built into aioson. No LLM required. No separate tool. Playwright is an optional runtime dependency — zero impact on users who don't need browser testing.
-- `qa:init`: generates `aios-qa.config.json` by reading `prd.md` (acceptance criteria → test scenarios) and `discovery.md` (business rules). URL auto-resolved from `project.context.md`.
-- `qa:doctor`: validates all prerequisites (Playwright installed, Chromium binary, config valid, URL reachable, context and prd.md present).
-- `qa:run`: full QA session with 4 personas: **naive** (empty forms, 10K strings, ghost clickables), **hacker** (8 secret patterns against window globals and HTML source, 10 sensitive file paths, XSS, open redirect, SQL injection, IDOR ±1, debug routes), **power** (keyboard navigation focus visibility, boundary values on numeric/date inputs), **mobile** (375px viewport, horizontal overflow, touch targets < 44px, fonts < 12px). Post-persona: network-level probes (console stack traces, sensitive GET params, mixed content). Accessibility audit (5 WCAG checks). Performance capture (TTFB, load time, request count, transfer size). AC coverage from `prd.md` with screenshots. Output: `aios-qa-report.md` + `aios-qa-report.json` + `aios-qa-screenshots/`.
-- `qa:scan`: autonomous crawler — maps all routes via BFS (configurable depth and max pages), probes each route for exposed secrets, console leaks, accessibility issues, and horizontal overflow. Sensitive files probed once per domain.
-- `qa:report`: displays last generated report; `--json` returns parsed `aios-qa-report.json`.
-- `@qa` agent (base + all 4 locales) updated with browser report integration rules: if `aios-qa-report.md` exists, merge findings, promote severity when both static review and browser test flag the same issue, tag ACs as `[browser-validated]`.
-- i18n keys added for `qa_doctor`, `qa_init`, `qa_run`, `qa_scan`, `qa_report` sections across all 4 locales (en, pt-BR, es, fr).
-- `docs/en/qa-browser.md`: full reference for all 5 qa: commands with examples, persona descriptions, probe list, performance thresholds, CI integration patterns.
-- README updated: Browser QA section in Commands, Agent usage helper, JSON output, and Docs feature guides — all with links to `docs/en/qa-browser.md`.
-
-## [0.1.25] - 2026-03-04
-### Added
-- `@product` agent: conversational product wizard that runs between `@setup` and `@analyst`. Starts from a raw idea and leads a natural back-and-forth conversation (8 strict conversation rules: one question at a time, no filler words, reflect before advancing, surface forgotten requirements, challenge assumptions gently, prioritize ruthlessly, draft early after 5–7 exchanges). Supports two modes: creation (no prd.md) and enrichment (prd.md exists). Produces `.aioson/context/prd.md` with 8 sections: vision, problem, users, MVP scope (🔴 must-have / 🟡 should-have), out-of-scope, user flows, success metrics, open questions.
-- `template/.aioson/agents/product.md`: base agent file.
-- `template/.aioson/locales/en/agents/product.md`, `pt-BR/agents/product.md`, `es/agents/product.md`, `fr/agents/product.md`: localized variants with absolute language instruction at top.
-- `template/.gemini/commands/aios-product.toml`: Gemini command registration.
-- Workflow chains updated in `config.md`: MICRO = `@setup → @product (optional) → @dev`; SMALL = `@setup → @product → @analyst → @architect → @dev → @qa`; MEDIUM = `@setup → @product → @analyst → @architect → @ux-ui → @pm → @orchestrator → @dev → @qa`.
-- `@product` added to `CLAUDE.md` agent routing, `AGENTS.md` invocation table and file list.
-- `setup.md` routing table updated in all 5 locale files to route to `@product` after setup instead of directly to `@analyst`.
-- `src/constants.js`: `@product` added to `MANAGED_FILES` (base + 4 locale paths + toml) and `AGENT_DEFINITIONS` (before `@analyst`, empty `dependsOn`, output `prd.md`).
-
-## [0.1.24] - 2026-03-04
-### Added
-- `skeleton-system.md`: new lightweight living index generated by `aioson-scan.py` alongside `discovery.md`. Contains file map with status indicators (✓/◑/○), key routes, module status table, and entity relationships. Designed to be read first as a cheap orientation before loading heavier context files.
-- Scanner now generates both `discovery.md` and `skeleton-system.md` in a single LLM call, split by `<<<SKELETON>>>` delimiter. If the delimiter is missing, only `discovery.md` is written with a warning.
-- `@dev` (base + en/pt-BR/es/fr): `skeleton-system.md` added to required input (read first). Auto-update rule: update skeleton entries when creating/deleting/modifying files. New `*update-skeleton` command to rewrite the full skeleton after a development session.
-- `@analyst` (base + en/pt-BR/es/fr): brownfield pre-flight now reads `skeleton-system.md` first for quick orientation before diving into `discovery.md`.
-- `@orchestrator` (base + en/pt-BR/es/fr): session start now reads `skeleton-system.md` as step 2 (before discovery.md).
-
-## [0.1.23] - 2026-03-04
-### Added
-- `template/aioson-scan.py`: standalone Python scanner (zero external dependencies, stdlib only). Walks the project tree, reads up to 12 key files (package.json, composer.json, schema.prisma, routes/web.php, etc.), reads `project.context.md` and `spec.md` if present, then calls a cheap LLM API (DeepSeek, Gemini, OpenAI, Anthropic, Groq, Together, or Mistral) to generate `.aioson/context/discovery.md` with 9 structured sections. Saves main-session tokens on brownfield projects.
-- `template/aioson-models.json`: API key config template for 7 providers. Auto-added to `.gitignore` on install to prevent key commits.
-- Brownfield project detection in installer: when `framework_installed` is detected and the project has >20 files, `install` and `update` now emit an alert with instructions to run `aioson-scan.py` first.
-- `@analyst` (base + en/pt-BR/es/fr): **Brownfield pre-flight** section — checks `framework_installed`, skips Phases 1–3 if `discovery.md` already exists, alerts the user to run the scanner if it is missing. Rule enforced: always read `spec.md` alongside `discovery.md`.
-- `@dev` (base + en/pt-BR/es/fr): **Brownfield alert** section — if `framework_installed=true` and `discovery.md` is missing, alerts the user to run `aioson-scan.py` before proceeding.
-- `@orchestrator` (base + en/pt-BR/es/fr): session start now reads `discovery.md` AND `spec.md` together at session open, with brownfield alert when `discovery.md` is absent.
-
-## [0.1.22] - 2026-03-04
-### Fixed
-- `setup:context` command: `--lang=pt-BR` (and any `--lang` alias) was silently ignored — `applyExplicitOverrides` only read `options.language` but the parser stores the flag as `options.lang`. Fixed by reading `options.language ?? options.lang`, consistent with all other commands (`install`, `update`, `init`, `locale-apply`, etc.). Running `npx aioson setup:context . --defaults --lang=pt-BR` now correctly installs the pt-BR locale agents.
-
-## [0.1.21] - 2026-03-03
-### Changed
-- `@ux-ui` Step 0 (base + en/pt-BR/es/fr locales): added **Option C — Default / Skip** to the visual style question. When chosen, the agent skips style confirmation and goes directly to Step 1 using `interface-design.md` as the sole design authority, letting domain exploration drive the visual direction organically. Accepted inputs: C / skip / pular / saltar / passer / padrão / default.
-
-## [0.1.20] - 2026-03-03
-### Added
-- `template/.aioson/skills/static/react-motion-patterns.md`: new skill with React/Next.js equivalents of the wow effects from `static-html-patterns.md`. Covers 10 patterns: animated mesh background, animated gradient text, scroll reveal (Framer Motion + Intersection Observer fallback), 3D card tilt hook, hero staggered entrance, infinite logo marquee, scroll progress bar, glassmorphism card, floating orbs, and page transition. All patterns include `prefers-reduced-motion` fallback and touch-device guards.
-- Laravel installation links to `skills/dynamic/laravel-docs.md`: three-path table (Herd/standard, Sail/Docker, Jetstream+Livewire) with specific URLs, detection signals per stack, and Livewire v3 MCP fetch example.
-
-### Changed
-- `template/.aioson/skills/static/laravel-conventions.md`: added project folder structure tree, naming convention table (13 artefacts, singular/plural rules), and Livewire component section with `#[Computed]`, `wire:model.live`, classic controller variant, and coexistence rule.
-- `@dev` agent (base + en/pt-BR/es/fr locales): added **Motion and animation** section — when `framework=React|Next.js`, agent now reads `react-motion-patterns.md` before implementing animations. Framer Motion as primary, CSS `@keyframes` as fallback. Hard rule against heavy motion in admin/CRUD interfaces.
-- `@dev` agent (base + en/pt-BR/es/fr locales): added Laravel project structure layout block and singular/plural naming rules to the Laravel conventions section.
-
-## [0.1.19] - 2026-03-03
-### Added
-- `src/commands/test-agents.js`: new `test:agents` CLI command — 99 structural checks across all 8 base agents and all locale files (32 base checks + 64 locale checks + 3 critical skill checks). Validates: file existence, `## Mission` section, `## Hard constraints` section, `.md-only` context rule, and language instruction within first 15 lines of each locale agent.
-- `src/commands/locale-diff.js`: new `locale:diff` CLI command — compares `##` heading structure of base agent files vs locale translations, normalizes headings (strips accents for cross-language comparison), and reports sections present in base but missing in locale. Supports `--lang` filter and single-agent argument.
-- `template/.aioson/skills/static/django-patterns.md`: comprehensive Django conventions skill (342 lines) — project structure, `TimestampMixin`, service layer, DRF serializers, URL namespacing, settings split (base/dev/prod), Admin registration, `pytest-django` test patterns, N+1 prevention rules.
-- `template/.aioson/skills/static/fastapi-patterns.md`: comprehensive FastAPI conventions skill (344 lines) — async project structure, `pydantic-settings`, SQLAlchemy async engine, Pydantic input/output schema separation, service layer, JWT via `HTTPBearer`, Alembic migrations, `pytest + httpx AsyncClient` patterns.
-- `template/.aioson/skills/dynamic/README.md`: explains static vs dynamic skills pattern and lists all 6 dynamic skill files with their source URLs.
-
-### Changed
-- `@dev` agent (base + en/pt-BR/es/fr locales): `architecture.md` and `discovery.md` now marked as `*(SMALL/MEDIUM only)*` in Required input. Added explicit MICRO callout — only `project.context.md` is guaranteed for MICRO projects.
-- `@orchestrator` base agent: renamed `## Rules` → `## Hard constraints` for consistency with all other agents.
-- `@dev`, `@qa`, `@orchestrator` base agents: added `.md-only` context rule (was missing — `test:agents` detected the gap).
-- Locale `en` agents (analyst, architect, pm, ux-ui, dev, qa, orchestrator): added ABSOLUTE INSTRUCTION language marker at top (was missing in en locale — `test:agents` detected the gap).
-- All 6 dynamic skill stubs expanded with concrete source URLs, focus areas, and MCP tool fetch examples to replace the minimal placeholder content.
-
-## [0.1.18] - 2026-03-03
-### Fixed
-- Language bug: locale agent files (pt-BR, es, fr) were responding in English even after `aioson install --lang=<locale>` because the language rule was at the bottom of each file. LLMs process top-to-bottom and would default to English before reading the rule.
-- Added `> **⚠ ABSOLUTE INSTRUCTION — LANGUAGE**` blockquote as the **first content after the title** in all 8 agents across all 3 locales (pt-BR, es, fr) + en locale (24 files total). Instruction explicitly forbids English and names the target language before any mission, sequence, or detection logic runs.
-- `@ux-ui` output contract: HTML deliverable was being saved to `.aioson/context/landing-preview.html` instead of the project root. Fixed to `index.html` in the project root across all 5 ux-ui files (base + en/pt-BR/es/fr locales).
-
-## [0.1.17] - 2026-03-03
-### Added
-- `static-html-patterns.md` Section 0 — **Hero Law**: explicit rule that the hero MUST be full-viewport animated background + ONE headline + TWO buttons. Cards in the hero are forbidden.
-- `static-html-patterns.md` Section 2a-extra — **Mandatory Wow Techniques** for Bold & Cinematic (three required, not optional):
-  - Animated mesh background (`@keyframes meshDrift 20s`) — static gradients replaced
-  - Animated gradient text (`@keyframes textGradient 8s`) on headline `<em>` key phrase
-  - 3D card tilt on hover (`perspective(700px) rotateX/rotateY` on `mousemove`, skipped on touch + reduced motion)
-- `@setup` agent (base + all 4 locales) — **Step 3: Next agent guidance** appended to Post-setup action. Agent now closes by explicitly naming the next `@agent` based on `project_type` + `classification`, using the exact `@name` format so AI clients (Codex, Claude Code, Gemini) can trigger it.
-- `@setup` pt-BR/es/fr/en locales — spec.md skip hint for `project_type=site` + MICRO classification.
-
-### Changed
-- `@ux-ui` Step 0 (base + en/pt-BR/es/fr locales) — upgraded from "ask one question" to **HARD STOP blocking gate**: agent must not read files, write HTML/CSS, or proceed to Step 1 until user answers the visual style question.
-- `@ux-ui` Landing page mode (base + all locales) — added **Hero Law** constraint and **Mandatory Wow Techniques** section explicitly referencing the three required animations.
-- `@ux-ui` es/fr locales — **full rewrite** to match the updated base agent: added Step 0 (visual style intake), landing page mode, hero law, mandatory wow techniques, full output contract.
-
-## [0.1.16] - 2026-03-03
-### Added
-- `template/.aioson/skills/static/static-html-patterns.md`: new **Section 14 — Premium Template Patterns (Aigocy-style)** with 10 production patterns extracted from a real AI-agency landing page (ThemeForest #61450410):
-  - **14a. effectFade animations**: `fadeUp` and `fadeRotateX` (3D perspective entrance) with `data-delay` stagger pattern for GSAP
-  - **14b. Infinite logo marquee**: CSS-only `@keyframes infiniteSlide` with auto-clone JS and hover-pause for accessibility
-  - **14c. SVG animated paths**: SMIL `<animateMotion>` hub-and-spoke diagram connecting icons to a center product image
-  - **14d. Scroll-to-top with circular progress**: CSS `stroke-dashoffset` progress ring updated by scroll JS
-  - **14e. Split Swiper**: synchronized text + image sliders with `effect: 'fade'` and `slideTo()` binding
-  - **14f. Swiper progress bar navigation**: thin animated fill bar replacing pagination dots for portfolio sliders
-  - **14g. box-white / box-black section alternation**: CSS pseudo-element radial glow replacing decorative PNG images
-  - **14h. Accordion FAQ**: native `<details>`/`<summary>` version with `rotate(45deg)` icon transition
-  - **14i. Footer with watermark background logo**: faded brand name in `position: absolute` behind 3-column dense footer
-  - **14j. Canvas cursor trail**: fading dot trail on `mousemove`, skipped on touch devices and `prefers-reduced-motion`
-- Section 13 pre-delivery checklist: added "No placeholder text remains" item
-
-### Changed
-- `@setup` agent output template (`aioson_version`) bumped to `0.1.16` across all locales (en, pt-BR, es, fr) and base agent file
-
-## [0.1.12] - 2026-03-02
-### Added
-- New static skill: `template/.aioson/skills/static/interface-design.md` — a comprehensive UI/UX craft guide derived from the interface-design project, covering: Intent-First Framework (3 mandatory questions before any layout), Domain Exploration (4 required outputs: domain concepts, color world, signature element, defaults to avoid), 6 design directions with full token specs (Precision & Density, Warmth & Approachability, Data & Analysis, Editorial, Commerce, Minimal & Calm), complete token architecture (foreground/background/border/brand/semantic roles), depth strategy (commit to ONE), component state matrix, 4 quality checks (swap/squint/signature/token tests), self-critique process (composition → craft → content → structure).
-- Portuguese documentation at `docs/pt/`:
-  - `README.md`: index linking all guides.
-  - `inicio-rapido.md`: quick start with install commands, classification scoring, and 3-command setup.
-  - `agentes.md`: per-agent reference with when-to-use, activation command, what it delivers, and concrete examples.
-  - `cenarios.md`: 4 complete worked examples — MICRO (landing page), SMALL (Laravel clinic API with @analyst output, @architect folder structure, @dev code), MEDIUM (Next.js SaaS with 3-lane parallel orchestration), MEDIUM dApp (Ethereum NFT marketplace with Solidity contract example).
-  - `web3.md`: Portuguese Web3 guide covering setup flags for Ethereum/Solana/Cardano, monorepo structure, per-agent Web3 conventions, and skill reference.
-
-### Changed
-- `@ux-ui` agent (base + all 4 locales: en/pt-BR/es/fr): mandatory pre-work now references `interface-design.md` skill, adds Intent-First step, Domain Exploration step (4 required outputs), single design direction declaration, 4 quality checks (swap/squint/signature/token tests), and self-critique gate before delivery. Output contract expanded with focal point, reading order, full state matrix, and handoff notes for signature visual moves.
-- `workflow:plan` sequences corrected: `@ux-ui` now included in SMALL (`setup → analyst → architect → ux-ui → dev → qa`) and MEDIUM (`setup → analyst → architect → ux-ui → pm → orchestrator → dev → qa`), matching `config.md` and agent documentation.
-- 13 static skills expanded from 5–7 stub lines to 200–337 lines of production-grade, code-first references:
-  - `laravel-conventions`: controllers as orchestrators, Form Requests, Actions, Policies, Events+queued Listeners, Jobs, API Resources, N+1 prevention, Model conventions, Migrations, Pest tests.
-  - `tall-stack-patterns`: Livewire lifecycle + real-time validation, inter-component events, lazy loading, Alpine.js scope rules, Tailwind design system discipline, full Flux UI examples (button/modal/table/dropdown/sidebar).
-  - `filament-patterns`: Resource structure (form/table/filters/actions), custom Pages, Stats Widgets, Relation Managers, policy enforcement, advanced form fields (repeater, conditional, file upload), business logic delegation.
-  - `flux-ui-components`: full component inventory, buttons (all variants + loading states), form field group, modal with programmatic control, dropdown menus, table with sort + empty state, badges, sidebar navigation.
-  - `jetstream-setup`: Inertia vs Blade decision matrix, Teams with roles/permissions, API tokens, profile customization, 2FA, password confirmation middleware, post-install checklist.
-  - `git-conventions`: full Conventional Commits spec with type table + examples, Git Flow vs GitHub Flow, branch naming, PR template, tagging, protected history rules.
-  - `nextjs-patterns`: Server vs Client mental model, App Router structure, async Server Components, Server Actions with Zod, Client Components (when/why), Route Handlers for webhooks only, metadata/SEO, loading.tsx + error.tsx.
-  - `node-express-patterns`: layered architecture (routes/controllers/services/repositories), Zod validation middleware, typed auth + role checking, AppError class hierarchy, centralized error handler, rate limiting, graceful shutdown.
-  - `node-typescript-patterns`: strict tsconfig baseline, Zod at all external boundaries, env validation at startup, branded domain types for IDs, const-object enums, repository interface pattern, explicit return types, asyncHandler.
-  - `rails-conventions`: Service Objects with Result type, model scopes/enums/validations, Active Record N+1 prevention, async Jobs with retry, Mailer patterns, serializers, Pundit authorization, RSpec request + unit specs.
-  - `web3-ethereum-patterns`: CEI pattern, ReentrancyGuard, pull over push, AccessControl roles, gas optimization (struct packing, custom errors), Hardhat test patterns with loadFixture, wagmi v2 frontend integration, deployment scripts.
-  - `web3-solana-patterns`: eUTxO/account model, Anchor program structure, account data with LEN, PDA seeds + bump storage, CPI signing, Anchor constraints, compute budget management, full Anchor test suite.
-  - `web3-cardano-patterns`: eUTxO mental model, Aiken project structure, datum/redeemer type design, spending validator, minting policy, Aiken tests, off-chain with Lucid, datum versioning strategy, deployment checklist.
-  - `web3-security-checklist`: 7 critical vulnerabilities with code examples (reentrancy, access control, integer overflow, oracle manipulation, flash loans, front-running, signature replay), pre-deployment checklist (static analysis/fuzzing/invariant tests/multisig/timelock), emergency response protocol.
-
-### Fixed
-- `workflow:plan` was silently omitting `@ux-ui` from SMALL and MEDIUM sequences despite `config.md` and all agent documentation specifying it as a required step. Fixed in `src/commands/workflow-plan.js` and updated `tests/workflow-plan.test.js`.
-
-## [0.1.11] - 2026-03-02
-### Added
-- Agent prompt enrichment across all 8 agents:
-  - `@analyst`: 6 concrete Phase 1 discovery questions, entity deep-dive example (scheduling system), field-level table format for Phase 3, `Visual references` and `Risks identified` output sections, responsibility boundary note.
-  - `@architect`: concrete folder/module structure trees for MICRO/SMALL/MEDIUM across Laravel (TALL), Node/Express, Next.js (App Router), and dApp (Hardhat/Foundry/Anchor) stacks.
-  - `@pm`: explicit 2-page golden rule with cut-ruthlessly instruction, when-to-use guidance (SMALL/MEDIUM only, skip MICRO), exact `prd.md` section template.
-  - `@dev`: Laravel ALWAYS/NEVER convention list (Form Requests, Actions, Policies, Events+Listeners, Jobs, Resources, N+1 prevention), UI/UX conventions, Web3 guards for dApp projects, semantic commit format with examples, responsibility boundary note.
-  - `@orchestrator`: MEDIUM-only activation condition with early exit, 4-step orchestration process, dependency graph example, parallel vs sequential classification rules, `agent-N.status.md` and `shared-decisions.md` status file protocol.
-  - `@setup`: explicit `framework_installed` contract semantics (true/false downstream behavior), monorepo detection guidance for mixed Web3 + backend repos.
-  - `@ux-ui` and `@qa`: no structural changes (already complete).
-- All 8 locale packs (`en`, `pt-BR`, `es`, `fr`) synchronized with enriched agent content.
-- `isMonorepoDetection()` in `src/detector.js`: returns `true` when a Web3 framework and a backend or frontend framework coexist in the same directory.
-- Monorepo detection note propagated to `setup:context` output (localized via `note_monorepo` key in all 4 i18n message files).
-- `note_monorepo` i18n key added to `en`, `pt-BR`, `es`, and `fr` message dictionaries.
-- 4 new tests for `isMonorepoDetection` in `tests/detector.test.js`.
-
-### Changed
-- `setup:context` now prepends a localized monorepo warning note when Web3 and application framework signals coexist in the project directory.
-- `template/.aioson/config.md` context contract updated with explicit `framework_installed` semantics.
-- `aioson_version` example in `@setup` output template corrected from `0.1.8` to `0.1.10`.
-- `tests/agent-contracts.test.js` updated to reflect new `@dev` section names (`Laravel conventions`, `Responsibility boundary`).
-
-### Added
-- Full `pt-BR` CLI dictionary at `src/i18n/messages/pt-BR.js`.
-- Localized agent prompt packs for:
-  - `es` at `template/.aioson/locales/es/agents/*.md`
-  - `fr` at `template/.aioson/locales/fr/agents/*.md`
-- New `@ux-ui` agent contract and template set:
-  - `.aioson/agents/ux-ui.md`
-  - `.aioson/locales/{en,pt-BR,es,fr}/agents/ux-ui.md`
-  - `.gemini/commands/aios-ux-ui.toml`
-- Legacy framework detection support:
-  - `CodeIgniter 3`
-  - `CodeIgniter 4`
-- New package installation validation command:
-  - `aioson test:package [source-path] [--keep] [--dry-run] [--json]`
-- New workflow planning command:
-  - `aioson workflow:plan [path] [--classification=MICRO|SMALL|MEDIUM] [--json]`
-- `mcp:init` tool presets generation under `.aioson/mcp/presets/`:
-  - `claude.json`
-  - `codex.json`
-  - `gemini.json`
-  - `opencode.json`
-- New MCP validation command:
-  - `aioson mcp:doctor [path] [--strict-env] [--json]`
-- New parallel orchestration bootstrap command:
-  - `aioson parallel:init [path] [--workers=2..6] [--force] [--dry-run] [--json]`
-- New parallel diagnosis/remediation command:
-  - `aioson parallel:doctor [path] [--workers=2..6] [--fix] [--force] [--dry-run] [--json]`
-- New parallel scope assignment command:
-  - `aioson parallel:assign [path] [--source=auto|prd|architecture|discovery|<file>] [--workers=2..6] [--force] [--dry-run] [--json]`
-- New parallel consolidated status command:
-  - `aioson parallel:status [path] [--json]`
-- Optional `mcp:init` tool filter:
-  - `--tool=claude|codex|gemini|opencode`
-- `test:smoke` mixed monorepo profile:
-  - `--profile=mixed` for combined Web2 + Web3 workspace validation
-- `test:smoke` parallel orchestration profile:
-  - `--profile=parallel` to validate `parallel:init/assign/status/doctor` in one flow
-- Formal JSON schema documentation for automation:
-  - `docs/en/json-schemas.md`
-  - `docs/en/schemas/index.json`
-  - `docs/en/schemas/*.schema.json` for each JSON command contract
-  - includes `docs/en/schemas/mcp-doctor.schema.json`
-  - includes `docs/en/schemas/parallel-init.schema.json`
-  - includes `docs/en/schemas/parallel-doctor.schema.json`
-  - includes `docs/en/schemas/parallel-assign.schema.json`
-  - includes `docs/en/schemas/parallel-status.schema.json`
-
-### Changed
-- Version resolution is now centralized via `src/version.js` and shared across `info`, `setup:context`, and installer metadata generation.
-- Removed remaining hardcoded `0.1.8` fallback in `setup:context` version assignment.
-- Locale resolution now supports canonical fallback for Portuguese:
-  - `pt-BR` -> `pt-br`
-  - `pt_br` -> `pt-br`
-  - `pt` -> `pt-br`
-- Locale resolution now also supports regional fallback for:
-  - `es-*` -> `es`
-  - `fr-*` -> `fr`
-- Agent locale resolution now supports:
-  - `es-*` -> `es`
-  - `fr-*` -> `fr`
-- `update` now reapplies active agent prompts from `conversation_language` in context, preventing locale reset after template refresh.
-- `mcp:init` now writes both shared plan and tool-specific preset templates.
-- `mcp:init` Context7/Database presets now default to remote-endpoint bridge templates (`mcp-remote`) with URL env vars instead of generic command placeholders.
-- `mcp:init` invalid `--tool` errors are now localized via i18n dictionaries.
-- `mcp:doctor` human-readable check messages and summary are now fully localized.
-- `workflow:plan` advisory notes are now localized from i18n keys instead of hardcoded English.
-- `setup:context` onboarding notes are now localized (including beginner recommendation notes and stack-option notes).
-- `test:package` failure messages are now localized via i18n dictionaries.
-- `parallel:doctor` check/hint messages are now fully localized (including severity labels).
-- Internal validation failures inside `test:smoke` are now localized instead of hardcoded English.
-- `parallel:status` human-readable status rows and lane summaries are now localized (including status labels).
-- `mcp:doctor` severity prefixes are now localized per locale (`OK/WARN/FAIL` equivalents).
-- `test:package` now localizes fallback command failure detail when stderr/stdout are empty.
-- `parallel:assign` lane scope summary lines are now localized in human-readable output.
-- Parallel commands now localize fallback `unknown` classification labels in human-readable errors.
-- `mcp:init` now avoids hardcoded fallback text for invalid `--tool` and uses i18n-backed messaging.
-- `mcp:init` now localizes generated server `reason` fields and preset `notes` content via i18n.
-- Removed remaining hardcoded `unknown` fallbacks in `parallel:doctor` check messages and `context:validate` parse-reason output.
-- `context:validate` now localizes known frontmatter parse reason codes into human-readable locale messages.
-- `mcp:doctor` now localizes context frontmatter parse reason codes using the same i18n mapping as `context:validate`.
-- Standardized localized line formatting for `agents`, `locale:apply`, `workflow:plan`, and `parallel:init` human-readable listings.
-- Localized diagnostic line wrappers for `mcp:doctor` and `parallel:doctor` checks/hints, plus `context:validate` issue list rows.
-- Localized diagnostic/action wrapper lines in `doctor` command output (checks, hints, fix actions, and detail lines).
-- `cli` help and unknown-command wrapper lines are now localized via i18n keys instead of inline formatting.
-- `init` and `install` now print explicit multi-IDE onboarding hints (`agents` + `agent:prompt setup`) with optional `--tool=codex|claude|gemini|opencode` guidance.
-- Agent flow now supports explicit UI/UX handoff:
-  - `@architect` may hand off key screens/component constraints to `@ux-ui`
-  - `@dev` now consumes `.aioson/context/ui-spec.md` when present
-  - `SMALL`/`MEDIUM` default sequence includes `@ux-ui`
-- `doctor` now enforces multi-IDE gateway contracts (Claude/Codex/Gemini/OpenCode) and validates required Gemini/OpenCode files.
-- `doctor` now also validates each Gemini command file (`.gemini/commands/aios-*.toml`) maps to the expected shared agent instruction file.
-- `doctor --fix` now restores broken gateway contract files (Claude/Codex/Gemini/OpenCode + Gemini command mappings) from template in safe mode.
-- `update` now supports `--lang=en|pt-BR|es|fr` (and `--language`) to force localized agent-pack sync during update, including dry-run planning.
-- `--json` output mode now supports `init`, `install`, and `update` with clean machine-readable payloads (no mixed human logs).
-- `--json` output mode now also supports `agents`, `agent:prompt`, `locale:apply`, `setup:context`, and `i18n:add`.
-- JSON payloads for `init/install/update/agents/agent:prompt/locale:apply/setup:context/i18n:add` now include stable `ok` and command context fields.
-- JSON schema catalog expanded with formal contracts for:
-  - `init`
-  - `install`
-  - `update`
-  - `agents`
-  - `agent:prompt`
-  - `locale:apply`
-  - `setup:context`
-  - `i18n:add`
-- `init` and `install` now support `--lang=en|pt-BR|es|fr` to auto-apply localized agent packs during bootstrap.
-- Setup templates now default `aioson_version` to `0.1.8`.
-- `ui-ux-modern` static skill was expanded to a production-ready checklist with token, state, accessibility, responsive, and handoff guidance.
-- Developer onboarding now handles `Other` backend/frontend choices as true free-text custom values (legacy/custom stacks).
-- Expanded automated coverage:
-  - `tests/i18n-cli.test.js`
-  - `tests/init-install-guidance.test.js`
-  - `tests/prompt-tool.test.js`
-  - `tests/init-install-guidance.test.js` now validates locale pack auto-apply on `init/install --lang`
-  - `tests/version.test.js`
-  - `tests/update.test.js` now covers `--lang` override and dry-run locale sync planning
-  - `tests/json-output.test.js` now covers `init/install/update --json`
-  - `tests/json-output.test.js` now covers `agents/agent:prompt/locale:apply/setup:context/i18n:add --json`
-  - `tests/json-schema-files.test.js` now validates expanded schema catalog
-  - `tests/locales.test.js`
-  - `tests/agent-contracts.test.js`
-  - `tests/smoke.test.js` now verifies active agent prompt locale application for `--lang=es` and `--lang=fr`
-  - `tests/mcp-doctor.test.js`
-  - `tests/mcp-init.test.js` now covers invalid `--tool` handling
-  - `tests/setup-context.test.js` now validates localized onboarding notes
-  - `tests/parallel-doctor.test.js` now validates localized check messages
-  - `tests/parallel-init.test.js`
-  - `tests/parallel-doctor.test.js`
-  - `tests/parallel-assign.test.js`
-  - `tests/parallel-status.test.js`
-  - `tests/json-output.test.js` now covers `mcp:doctor --json`
-  - `tests/package-test.test.js`
-  - `tests/workflow-plan.test.js`
-  - extended `tests/mcp-init.test.js` and `tests/json-output.test.js`
-  - smoke JSON e2e checks for `--locale=es` and `--locale=fr`
-  - smoke coverage for `--profile=parallel`
-
-## [0.1.7] - 2026-03-01
-### Added
-- JSON output mode (`--json`) for:
-  - `aioson info`
-  - `aioson doctor`
-  - `aioson context:validate`
-  - `aioson test:smoke`
-- New JSON output test suite: `tests/json-output.test.js`.
-
-### Changed
-- CLI now returns structured JSON errors for unknown commands and runtime failures when `--json` is enabled.
-- `setup:context` and setup templates now default `aioson_version` to `0.1.7`.
-
-## [0.1.6] - 2026-03-01
-### Added
-- `test:smoke` now supports chain-specific Web3 profiles:
-  - `--web3=ethereum`
-  - `--web3=solana`
-  - `--web3=cardano`
-- Web3 smoke workflow now verifies:
-  - framework detection per chain profile
-  - `project.context.md` dApp/Web3 frontmatter consistency.
-
-### Changed
-- CLI help and docs updated for `test:smoke --web3`.
-- `setup:context` and setup templates now default `aioson_version` to `0.1.6`.
-
-## [0.1.5] - 2026-03-01
-### Added
-- Web3 framework detection:
-  - Ethereum: `Hardhat`, `Foundry`, `Truffle`
-  - Solana: `Anchor`, `Solana Web3`
-  - Cardano: `Cardano` (Aiken/Cardano SDK signals)
-- New Web3 skill templates:
-  - static: `web3-ethereum-patterns`, `web3-solana-patterns`, `web3-cardano-patterns`, `web3-security-checklist`, `node-typescript-patterns`
-  - dynamic: `ethereum-docs`, `solana-docs`, `cardano-docs`
-- New documentation page: `docs/en/web3.md`.
-
-### Changed
-- `project_type` now accepts `dapp`.
-- `setup:context` now supports Web3 context fields (`web3_enabled`, `web3_networks`, `contract_framework`, `wallet_provider`, `indexer`, `rpc_provider`).
-- `setup:context` and setup templates now default `aioson_version` to `0.1.5`.
-
-## [0.1.4] - 2026-03-01
-### Added
-- New command:
-  - `aioson test:smoke [workspace-path] [--lang=en|pt-BR] [--keep]`
-- New smoke test suite: `tests/smoke.test.js`.
-
-### Changed
-- `doctor` now supports safe remediation mode:
-  - `aioson doctor --fix`
-  - `aioson doctor --fix --dry-run`
-- `setup:context` and setup templates now default `aioson_version` to `0.1.4`.
-
-## [0.1.3] - 2026-03-01
-### Added
-- Localized agent prompt packs:
-  - `.aioson/locales/en/agents/*.md`
-  - `.aioson/locales/pt-BR/agents/*.md`
-- New command:
-  - `aioson locale:apply [path] [--lang=en|pt-BR] [--dry-run]`
-- Agent path resolution now supports locale-aware prompts with fallback to active agent paths.
-- New locale test suite: `tests/locales.test.js`.
-
-### Changed
-- `setup:context` now applies localized agent prompts based on `conversation_language`.
-- Documentation updated for locale pack workflow.
-
-## [0.1.2] - 2026-03-01
-### Added
-- New commands:
-  - `aioson agents`
-  - `aioson agent:prompt <agent> [--tool=...]`
-  - `aioson context:validate [path]`
-  - `aioson setup:context [path]` (interactive or defaults mode)
-- New modules:
-  - `src/context.js` for frontmatter parsing and context contract validation
-  - `src/agents.js` for agent metadata and prompt generation
-  - `src/context-writer.js` for context rendering and classification scoring
-- New test suites:
-  - `tests/context.test.js`
-  - `tests/agents.test.js`
-  - `tests/context-writer.test.js`
-
-### Changed
-- `doctor` now validates `project.context.md` frontmatter and required fields, including `conversation_language` format.
-
-## [0.1.1] - 2026-03-01
-### Changed
-- Converted remaining template files to English (`.aioson` agents/config and Gemini command descriptions).
-- Hardened `@setup` instructions to prevent silent defaults for `project_type`, `profile`, `classification`, and language.
-- Added explicit `conversation_language` to context contract so agent interaction language can be enforced.
-
-## [0.1.0] - 2026-03-01
-### Added
-- Initial CLI commands: init, install, update, info, doctor
-- Multi-IDE template gateways (Claude, Codex, Gemini, OpenCode)
-- Framework detector and installer/updater core
-- i18n message system with English default
-- Automated tests for detector, installer, doctor, i18n
-- `i18n:add <locale>` command to scaffold new locale dictionaries
-- GitHub Actions CI and tag-based npm release workflows
-
-### Changed
-- Project-facing content standardized to English
-- CLI i18n upgraded with dynamic locale loading and fallback behavior
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+## [Unreleased]
+
+### Added
+- **@pentester agent**: adversarial security review agent with structured findings output (`security-findings-{slug}.json`) and Gate D blocking capability for MEDIUM projects.
+- **@discover agent**: system discovery and semantic knowledge cache bootstrap for brownfield projects.
+- **git:guard**: pre-commit guardrail that blocks forbidden files (`node_modules/`, secrets, build artifacts) from being staged or committed. Supports `--install-hook` for persistent protection.
+- **commit:prepare**: automated commit preparation command that collects staged diffs, runs `git:guard`, and generates `commit-prep.json` ready for `@committer`.
+- **compress:agents**: token-reduction command with structural (free) and semantic (`--llm`) modes. Backs up originals to `.original.md` and supports `--restore`.
+- **tmux launcher**: `live:start` now supports tmux for persistent terminal sessions with compact ANSI status bars.
+- **Runner system**: `runner:run`, `runner:queue`, `runner:plan`, and `runner:daemon` commands for persistent background job execution.
+- **Design-docs governance**: modular code governance system with 5 best-practice files (`folder-structure`, `componentization`, `code-reuse`, `naming`, `file-size`) distributed automatically on install and loaded unconditionally by `@dev` and `@deyvin`.
+- **SDLC process upgrade**: gates and handoff protocol enhancements across the workflow engine.
+- **Feature closure automation**: `feature:close` now auto-triggers `feature:archive` on `--verdict=PASS`, moving all feature artefacts (`prd-`, `spec-`, `requirements-`, etc.) to `.aioson/context/done/{slug}/` and updating `done/MANIFEST.md`. No manual steps required.
+- **Feature archive command**: `feature:archive` with `--dry-run`, `--restore`, and `--force` for retrospective archival and archive restoration. Replaces the legacy `context:trim` workflow.
+- **Agent awareness of archived features**: `@cypher`, `@discover`, `@neo`, and `@sheldon` now read `.aioson/context/done/MANIFEST.md` to avoid loading full archived files.
+
+### Changed
+- **i18n architecture**: decoupled interaction language from agent prompts. Agent instruction files are now canonical English only; `conversation_language` in `project.context.md` controls user-facing language. Removed localized agent packs in favor of single canonical source.
+- **Agent manifests**: moved `.manifest.json` files to `.aioson/agents/manifests/` subfolder to reduce clutter.
+- **Core agents refactored**: `@product`, `@sheldon`, `@dev`, `@deyvin`, `@ux-ui`, and `@squad` split into deterministic on-demand modules for better context efficiency.
+- **@committer**: enhanced with terminal checkbox UI, robust prepare fallback, and optimized workflow.
+- **@squad**: enhanced with domain classification gate, investigation handoff, language policy, and package contract restoration.
+- **Installer pipeline**: hardened with pentester agent integration and improved integrity checks.
+
+### Fixed
+- `@dev` pt-BR locale realignment with canonical prompt flow.
+- `@squad` genome bindings and package contract restoration.
+- Legacy process safeguards restored across agents.
+- Safe canonical English agent sources restored after i18n decoupling.
+- Accidentally tracked local directories removed from git tracking.
+
+## [1.7.3] - 2026-04-13
+### Fixed
+- `@dev` pt-BR locale pack realigned with the canonical prompt flow, restoring the proper cold-start fallback when `dev-state.md` is already `done` and another feature is still `in_progress` in `features.md`.
+- Added a regression test to keep the `@dev` prompt synchronized between the canonical file and the pt-BR locale pack, preventing future drift in the distributed template.
+- Setup prompt templates now reference the current `aioson_version` (`1.7.3`) instead of the stale `1.5.1` example.
+
+## [1.7.2] - 2026-04-07
+### Fixed
+- Removed 5 spurious Windows alternate data stream files (`SKILL.md:Zone.Identifier`) from `template/.aioson/skills/design-system/` that leaked from a local Windows copy operation and were never tracked by git.
+- `tests/agent-contracts.test.js`: added `neo`, `sheldon`, `tester`, and `orache` to the full locale coverage check; added a separate `BASE_ONLY_AGENTS` list (`copywriter`, `design-hybrid-forge`, `site-forge`) with a dedicated test that verifies their base files ship in the template. Previously these agents were not validated by the test suite.
+
+## [1.7.1] - 2026-04-07
+### Added
+- **Marketing & copywriting system**: new `@copywriter` agent (5 operating modes including VSL scripts) backed by a generic `copywriting` genome (One Belief, 5 Acts, PMS framework, market sophistication levels, 10 heuristics).
+- **Marketing references library** (`.aioson/skills/marketing/references/`): 8 on-demand reference files with 70+ patterns and anti-patterns covering One Belief, 5-Act narrative structure, fascinations (12 formulas), offer structure, PMS research, copy patterns, anti-patterns checklist, and market intelligence tools. Loaded conditionally by `@copywriter` to keep context lean.
+- **VSL Craft skill** (`.aioson/skills/marketing/vsl-craft.md`): Video Sales Letter production playbook — 3 formats (horizontal/vertical/hybrid), 5-act script structure, 5 hook formulas, retention techniques, testing methodology, and production specs.
+- **Marketing integration across agents**: `@neo` routes to `@copywriter` when `project_type=site` and copy is missing; `@setup` registers the copywriting genome and adds copy gate to site routing; `@dev` enforces a copy guard for marketing pages and reads the 5-Act structure from `copy-{slug}.md`; `@ux-ui` maps the 5 Acts to visual sections; `@deyvin` routes marketing copy requests to `@copywriter` instead of writing inline.
+- **Landing page production skills**: `landing-page-forge.md` (animation libraries, performance, SEO/LLMO, tracking, 3-track parallel production checklist) and `landing-page-deploy.md` (Vercel and Hostinger VPS deploy paths).
+- **Digital marketing agency squad template** under `template/.aioson/templates/squads/digital-marketing-agency/`.
+
+### Changed
+- `@dev` site production section now includes a copy guard that blocks inline marketing copy on `project_type=site` pages — copy must come from `@copywriter` via `copy-{slug}.md`.
+- `@ux-ui` copy gate now reads the 5-Act narrative structure from copy files and maps acts to UI sections.
+- `@neo` workflow stage detection includes `Needs copy` for site projects without a copy file.
+- `@setup` routing table splits `site` into two paths: `@copywriter → @ux-ui → @dev` (no copy yet) vs `@ux-ui → @dev` (copy ready).
+- Misc updates to `@qa`, `@squad`, `squad-create`, `squad-design`, `cognitive-core-ui/motion`, `ui-ux-modern`, and `template/CLAUDE.md`.
+
+### Notes
+- Marketing references are loaded conditionally by `@copywriter` only — other agents never load them directly. This keeps the marketing system isolated and prevents contaminating non-site projects.
+- The copy guard triggers exclusively for `project_type=site` marketing/sales pages. SaaS UI, dashboards, app interfaces, and documentation sites use the standard implementation flow with no behavior change.
+
+## [1.7.0] - 2026-04-06
+### Added
+- **Squad autonomous execution system**: full multi-agent squad stack — intra-squad message bus (`intra-bus.js`, `squad:bus` command), executor reflection module (`reflection.js`), task decomposer, `squad:autorun` for end-to-end autonomous execution, learning system, squad daemon, verify-gate, and cross-AI synthesis. Intelligence gaps addressed in plans 80–82; 6 critical operational fixes in plan 83.
+- **Runner system**: `runner:run`, `runner:queue`, `runner:plan`, and `runner:daemon` commands for persistent background job execution outside the main session loop.
+- **New CLI commands**: `agent:audit` (agent integrity inspection), `brief:gen` (context brief generation), and `verify:gate` (execution pre-flight gate). Documented in `docs/pt/`.
+- **SDD automation — The 80% Rule (plans 74–79)**: automation scripts that drive spec compliance across agents; harness sensors, context budget ceiling, and PGE pattern (plan 76); full SDD coverage across all agents (plan 77); locale SDD sync and pt-BR `tester.md` (plan 78); SDD automation scripts wired into agents and docs (plan 79); engine fixes for `evaluateReadiness`, `extractLastCheckpoint`, and `detectTestRunner`.
+- **Agent capability sprints**:
+  - Sprint 1 — task-list working memory, hook contract, file size guidelines, `CLAUDE.local.md` support.
+  - Sprint 2 — context compaction protocol and self-directed plan mode.
+  - Sprint 3 — CronTools protocol and config tiers documentation.
+- **Web-research-cache skill**: caches web research results in `researchs/`; awareness propagated to all agents that perform web discovery.
+- **`site-forge` agent**: replaces `hybrid-clone` with a dedicated site-forge agent backed by a Brains knowledge system for site cloning and forging workflows.
+- **Three.js skill**: `threejs-spatial` modifier and corresponding agent (`hybrid-clone`) for Three.js spatial interface projects.
+- **Operational pipeline**: event enrichment, spec-sync, token economy, devlog pipeline, and project hooks system.
+- **Squad CLI integration**: `squad.md` wired to CLI tools — Step 0 scaffold, CLI integration table.
+- **Tutorials site**: squads tutorial page added; Squads and Automation links activated on the tutorials index.
+
+### Changed
+- `update` now only updates already-installed files by default; use `--all` to sync every template file.
+- Setup routing no longer directs to `@dev` when only `plans/` or research files are present.
+- `@deyvin`, `@neo`, and `@sheldon` now load and maintain `dev-state.md` awareness.
+- README redesigned with visual impact, squads ecosystem section, and spec-driven workflow documentation.
+
+### Fixed
+- `@orache` research cache: agent now writes to `researchs/` (previously only read from it).
+- SDD engine reliability: `evaluateReadiness`, `extractLastCheckpoint`, `detectTestRunner` (plan 79 phase 2).
+- Squad autonomous operation: 6 critical fixes for stable real-world execution (plan 83).
+- Three.js patterns: race condition and invalid property reference resolved.
+
+## [1.6.0] - 2026-04-01
+### Added
+- **Interactive install wizard**: animated multi-screen onboarding (`init` and `install`) with Design (screen 3) and Locale (screen 4) dimensions, multi-select for design skills, and pre-populated choices on `--reconfigure`.
+- **`aioson setup` command**: unified setup entry point with auto-detected system language.
+- **Spec-driven process skill** (`aioson-spec-driven`): agent contract upgrades, GSD lessons integration (must_haves, 4-tier verification, gray areas, seeds, forensics), and context budget warning threshold configuration.
+- **Design skills**: `glassmorphism-ui` and `neo-brutalist-ui` added; HTML preview gallery for all 8 design skills with per-skill landing page previews.
+- **Aurora Command UI hybrid**: design hybrid variation workflow and aurora-command-ui skill.
+- **5-phase context optimization system**: implemented in `src/commands/context-optimizations.js`.
+- **Health digest, dynamic tools, and learning evolve pipeline**: `--all` flag for the main update flow; health digest output and adaptive learning pipeline wired to the operational surface.
+
+### Changed
+- `install --reconfigure`: warns when the new profile deselects previously installed items; confirm screen strings fully localized.
+- `update`: now preserves locale from the saved profile and installs all framework files when upgrading; added `--all` flag.
+
+### Fixed
+- `install`: `overwrite=true` enforced when reconfigure changes the active profile.
+- `design-hybrid` options: TTY error logged correctly; test coverage expanded.
+- `install-wizard`: broken banner box, wrong skip reason, and stdin hang after wizard completion resolved; terminal state now fully restored.
+- Landing page previews: hero content and section titles centered correctly across all 3 previews.
+
+## [1.5.1] - 2026-03-28
+### Added
+- **TDD Gate no `@dev`**: novo bloqueador explícito antes de qualquer implementação de lógica de negócio. Detecta test runner via `test_runner` em `project.context.md` ou varre a raiz em busca de `pest.xml`, `vitest.config.*`, `pytest.ini`, `.rspec`, `foundry.toml`. Mandato RED → GREEN → commit por classificação (MICRO/SMALL/MEDIUM) com exceções para migrations, configurações e conteúdo estático.
+- **Novo agente `@tester`**: engenheiro de testes para aplicações já implementadas. Produz `test-inventory.md` (mapa de cobertura), `test-plan.md` (estratégia escolhida + cobertura antes/depois) e escreve testes por módulo em ordem de risco. Suporta Characterization Testing, Test Pyramid, Risk-first Gap Filling, Property-based, Contract e Mutation Testing. Mapeamento completo de frameworks: Laravel/Pest, Next.js/Vitest+RTL, Django/pytest, FastAPI/httpx, Rails/RSpec, Solidity/Foundry.
+- **Detecção de documentos de kickoff no `@product` e `@sheldon`**: ambos os agentes agora varrem `plans/*.md` e `prds/*.md` na raiz do projeto antes de iniciar o fluxo. Detectam contexto (greenfield vs feature), oferecem sintetizar o conteúdo nos artefatos formais em `.aioson/context/`, e nunca modificam os arquivos fonte — o usuário decide quando deletá-los.
+- Campo `test_runner` adicionado ao Context contract em `config.md` e ao template de `project.context.md` gerado pelo `@setup` (todos os 4 locales).
+- Detecção automática de test runner no `@setup`: varre a raiz e preenche `test_runner` em `project.context.md` quando um runner é detectado.
+- `@tester` registrado em `CLAUDE.md`, `AGENTS.md`, `OPENCODE.md`, `.gemini/commands/aios-tester.toml`, `src/constants.js` (AGENT_DEFINITIONS) e `template/.claude/commands/aioson/agent/tester.md`.
+- Documentação `docs/pt` atualizada: `agentes.md` (seções `@dev`, `@qa`, `@product`, `@sheldon` novo, `@tester` novo, fluxos atualizados), `cenarios.md` (exemplo TDD no Cenário 2, exemplo kickoff doc, nota @tester vs @qa), `inicio-rapido.md` (tabela de sequências e seção de estratégia de kickoff docs).
+
+### Changed
+- `src/context-writer.js` e `src/commands/setup-context.js`: suporte ao campo `testRunner` na geração e override de `project.context.md`.
+- `template/.aioson/agents/dev.md`: seção "Implementation strategy" reescrita com mandato RED→GREEN→commit; seção "Atomic execution" atualizada com referência cruzada ao TDD Gate.
+- `template/.aioson/agents/product.md`: seção "Source document detection" adicionada antes de "Mode detection".
+- `template/.aioson/agents/sheldon.md`: seção "Detecção de documentos fonte" adicionada antes de RF-01.
+- `template/.aioson/config.md`: campo `test_runner` no Context contract; `@tester` no workflow opcional.
+- Versão de referência nos templates de `project.context.md` atualizada de `0.1.25` para `1.5.1`.
+
+## [1.5.0] - 2026-03-27
+### Added
+- Native web map and scrape commands for first-party web discovery and extraction workflows.
+- New agent capabilities and orchestration surface including `@neo`, the Sheldon PRD enrichment flow, richer downstream gates, and improved locale-aware runtime execution rules.
+- Squad-level operational improvements including webhook/channel bridge integration, daemon and registry evolution, and dashboard/design-system refinements for the squad experience.
+- Cognitive Core UI design skill upgrades with stronger art direction, composition libraries, anti-generic guardrails, and more robust interaction/contrast guidance.
+
+### Changed
+- Project license changed from `MIT` to `AGPL-3.0-only` for new releases going forward.
+- Package metadata and lockfile versioning are now aligned at `1.5.0`.
+- Agent workflows now enforce PRD gating, stronger atomic execution behavior, and more explicit routing for MICRO and product/design-driven flows.
+
+## [1.3.0] - 2026-03-15
+### Added
+- Canonical runtime observability for installed projects: runtime SQLite now acts as the primary execution store with task/run/event lifecycle tracking, workflow event synchronization, and an execution gateway prepared to instrument direct agent entrypoints, workflow enforcement, squads, and operational runs from the project runtime itself.
+- `docs/pt/comandos-cli.md`: first Portuguese-first CLI reference page covering all public `aioson` commands with grouped descriptions, command map, examples, and practical usage recipes. Linked from `docs/pt/README.md` and `docs/pt/inicio-rapido.md`.
+
+### Changed
+- Full product rebrand from `aioson` to `aioson` across package metadata, CLI branding, docs, templates, prompts, tests, runtime paths, and generated project structure. The managed project directory is now `.aioson/` instead of `.aioson/`.
+- npm distribution moved to the scoped package `@jaimevalasek/aioson` after the unscoped `aioson` name was rejected by npm for similarity policy reasons. Install and `npx` docs now use the scoped package while the installed CLI commands remain `aioson` and `aios`.
+- Release description updated to `AI operating framework for hyper-personalized software.` to match the current platform direction.
+- Legacy `dashboard:*` CLI flow removed from the public surface. `aioson dashboard:init|dev|open` no longer appear in help or docs; if called, the CLI now returns a migration error instructing users to open the separately installed dashboard app and select the project folder that already contains `.aioson/`. Updated pt docs and agent guidance to match the new app-first dashboard flow.
+- `template/.aioson/skills/static/interface-design.md`: upgraded the interface-design craft guide by merging strengths from the upstream `interface-design` skill without losing AIOSON workflow fit. Added persistent design memory via `.interface-design/system.md`, existing-UI refinement rules (diagnose drift before re-theme), two fully specified directions (`Sophistication & Trust`, `Premium Dark Platform`), a mandatory pre-component decision checkpoint, and explicit memory update guidance for reusable UI patterns.
+
+## [0.1.28] - 2026-03-04
+### Added
+- **`scan:project` command** (`aioson scan:project [path] [--provider=<name>] [--dry-run] [--json]`): Node.js port of the brownfield project scanner, replacing `aioson-scan.py`. Walks the project, reads key files, calls a cheap LLM (multi-provider), and generates `discovery.md` + `skeleton-system.md`. Zero npm dependencies — uses `node:fs`, `node:https`, `node:http` only.
+- Multi-provider HTTP client supporting OpenAI-compatible format (DeepSeek, OpenAI, Gemini, Groq, Together, Mistral) and Anthropic native format. Configured via `aioson-models.json` (same config file as before).
+- `--provider=<name>` flag to override `preferred_scan_provider` from config at runtime.
+- `--dry-run` mode: walks the project and reports file counts without calling the LLM.
+- i18n keys for `scan_project` section in all 4 locales (en, pt-BR, es, fr).
+- `install` hint message updated in all 4 locales: `python aioson-scan.py` → `aioson scan:project`.
+
+### Removed
+- `template/aioson-scan.py`: Python scanner removed from the template. `aioson scan:project` is the normalized replacement. The `aioson-models.json` config file (provider API keys) is unchanged.
+
+### Changed
+- All 15 template agent files (analyst, dev, orchestrator — base + 4 locales each): `python aioson-scan.py` replaced with `aioson scan:project` in brownfield detection hints.
+
+## [0.1.27] - 2026-03-04
+### Added
+- **HTML reports** for browser QA: `--html` flag added to `qa:run`, `qa:scan`, and `qa:report`. Generates a self-contained visual report in `reports/<date>_<time>_<mode>/index.html`. Screenshots embedded as base64 — fully portable for sharing. Existing `aios-qa-report.md` and `aios-qa-report.json` outputs are never modified; HTML is purely additive.
+- `reports/index.html`: historical index auto-updated after each HTML run, listing all runs with date, mode, URL, and severity badges. Sorted newest first.
+- `qa:report --html`: retroactively generates HTML from the existing `aios-qa-report.json` without re-running Playwright.
+- HTML report features: severity filter buttons, collapsible finding cards (location + risk + fix + screenshot), performance cards with colour-coded thresholds, AC coverage table, routes list (scan mode), click-to-zoom screenshot lightbox, vanilla JS (no framework).
+- `src/qa-html-report.js`: self-contained HTML generator module with embedded CSS and JS, `writeHtmlReport()` and `updateReportsIndex()` exports.
+- i18n key `html_report_written` added to `qa_run`, `qa_scan`, `qa_report` sections in all 4 locales (en, pt-BR, es, fr).
+- Missing `help_qa_*` i18n keys added to pt-BR locale (were absent; es/fr fall back to en).
+- `docs/en/qa-browser.md`: `--html` flag documented for all three commands, plus a new **HTML reports** section describing folder structure, features, and `qa:report --html` retroactive mode.
+- README: `--html` flag added to Browser QA command signatures and Agent usage helper examples.
+
+## [0.1.26] - 2026-03-04
+### Added
+- **Browser QA engine** (`qa:init`, `qa:doctor`, `qa:run`, `qa:scan`, `qa:report`): full Playwright-powered browser testing built into aioson. No LLM required. No separate tool. Playwright is an optional runtime dependency — zero impact on users who don't need browser testing.
+- `qa:init`: generates `aios-qa.config.json` by reading `prd.md` (acceptance criteria → test scenarios) and `discovery.md` (business rules). URL auto-resolved from `project.context.md`.
+- `qa:doctor`: validates all prerequisites (Playwright installed, Chromium binary, config valid, URL reachable, context and prd.md present).
+- `qa:run`: full QA session with 4 personas: **naive** (empty forms, 10K strings, ghost clickables), **hacker** (8 secret patterns against window globals and HTML source, 10 sensitive file paths, XSS, open redirect, SQL injection, IDOR ±1, debug routes), **power** (keyboard navigation focus visibility, boundary values on numeric/date inputs), **mobile** (375px viewport, horizontal overflow, touch targets < 44px, fonts < 12px). Post-persona: network-level probes (console stack traces, sensitive GET params, mixed content). Accessibility audit (5 WCAG checks). Performance capture (TTFB, load time, request count, transfer size). AC coverage from `prd.md` with screenshots. Output: `aios-qa-report.md` + `aios-qa-report.json` + `aios-qa-screenshots/`.
+- `qa:scan`: autonomous crawler — maps all routes via BFS (configurable depth and max pages), probes each route for exposed secrets, console leaks, accessibility issues, and horizontal overflow. Sensitive files probed once per domain.
+- `qa:report`: displays last generated report; `--json` returns parsed `aios-qa-report.json`.
+- `@qa` agent (base + all 4 locales) updated with browser report integration rules: if `aios-qa-report.md` exists, merge findings, promote severity when both static review and browser test flag the same issue, tag ACs as `[browser-validated]`.
+- i18n keys added for `qa_doctor`, `qa_init`, `qa_run`, `qa_scan`, `qa_report` sections across all 4 locales (en, pt-BR, es, fr).
+- `docs/en/qa-browser.md`: full reference for all 5 qa: commands with examples, persona descriptions, probe list, performance thresholds, CI integration patterns.
+- README updated: Browser QA section in Commands, Agent usage helper, JSON output, and Docs feature guides — all with links to `docs/en/qa-browser.md`.
+
+## [0.1.25] - 2026-03-04
+### Added
+- `@product` agent: conversational product wizard that runs between `@setup` and `@analyst`. Starts from a raw idea and leads a natural back-and-forth conversation (8 strict conversation rules: one question at a time, no filler words, reflect before advancing, surface forgotten requirements, challenge assumptions gently, prioritize ruthlessly, draft early after 5–7 exchanges). Supports two modes: creation (no prd.md) and enrichment (prd.md exists). Produces `.aioson/context/prd.md` with 8 sections: vision, problem, users, MVP scope (🔴 must-have / 🟡 should-have), out-of-scope, user flows, success metrics, open questions.
+- `template/.aioson/agents/product.md`: base agent file.
+- `template/.aioson/locales/en/agents/product.md`, `pt-BR/agents/product.md`, `es/agents/product.md`, `fr/agents/product.md`: localized variants with absolute language instruction at top.
+- `template/.gemini/commands/aios-product.toml`: Gemini command registration.
+- Workflow chains updated in `config.md`: MICRO = `@setup → @product (optional) → @dev`; SMALL = `@setup → @product → @analyst → @architect → @dev → @qa`; MEDIUM = `@setup → @product → @analyst → @architect → @ux-ui → @pm → @orchestrator → @dev → @qa`.
+- `@product` added to `CLAUDE.md` agent routing, `AGENTS.md` invocation table and file list.
+- `setup.md` routing table updated in all 5 locale files to route to `@product` after setup instead of directly to `@analyst`.
+- `src/constants.js`: `@product` added to `MANAGED_FILES` (base + 4 locale paths + toml) and `AGENT_DEFINITIONS` (before `@analyst`, empty `dependsOn`, output `prd.md`).
+
+## [0.1.24] - 2026-03-04
+### Added
+- `skeleton-system.md`: new lightweight living index generated by `aioson-scan.py` alongside `discovery.md`. Contains file map with status indicators (✓/◑/○), key routes, module status table, and entity relationships. Designed to be read first as a cheap orientation before loading heavier context files.
+- Scanner now generates both `discovery.md` and `skeleton-system.md` in a single LLM call, split by `<<<SKELETON>>>` delimiter. If the delimiter is missing, only `discovery.md` is written with a warning.
+- `@dev` (base + en/pt-BR/es/fr): `skeleton-system.md` added to required input (read first). Auto-update rule: update skeleton entries when creating/deleting/modifying files. New `*update-skeleton` command to rewrite the full skeleton after a development session.
+- `@analyst` (base + en/pt-BR/es/fr): brownfield pre-flight now reads `skeleton-system.md` first for quick orientation before diving into `discovery.md`.
+- `@orchestrator` (base + en/pt-BR/es/fr): session start now reads `skeleton-system.md` as step 2 (before discovery.md).
+
+## [0.1.23] - 2026-03-04
+### Added
+- `template/aioson-scan.py`: standalone Python scanner (zero external dependencies, stdlib only). Walks the project tree, reads up to 12 key files (package.json, composer.json, schema.prisma, routes/web.php, etc.), reads `project.context.md` and `spec.md` if present, then calls a cheap LLM API (DeepSeek, Gemini, OpenAI, Anthropic, Groq, Together, or Mistral) to generate `.aioson/context/discovery.md` with 9 structured sections. Saves main-session tokens on brownfield projects.
+- `template/aioson-models.json`: API key config template for 7 providers. Auto-added to `.gitignore` on install to prevent key commits.
+- Brownfield project detection in installer: when `framework_installed` is detected and the project has >20 files, `install` and `update` now emit an alert with instructions to run `aioson-scan.py` first.
+- `@analyst` (base + en/pt-BR/es/fr): **Brownfield pre-flight** section — checks `framework_installed`, skips Phases 1–3 if `discovery.md` already exists, alerts the user to run the scanner if it is missing. Rule enforced: always read `spec.md` alongside `discovery.md`.
+- `@dev` (base + en/pt-BR/es/fr): **Brownfield alert** section — if `framework_installed=true` and `discovery.md` is missing, alerts the user to run `aioson-scan.py` before proceeding.
+- `@orchestrator` (base + en/pt-BR/es/fr): session start now reads `discovery.md` AND `spec.md` together at session open, with brownfield alert when `discovery.md` is absent.
+
+## [0.1.22] - 2026-03-04
+### Fixed
+- `setup:context` command: `--lang=pt-BR` (and any `--lang` alias) was silently ignored — `applyExplicitOverrides` only read `options.language` but the parser stores the flag as `options.lang`. Fixed by reading `options.language ?? options.lang`, consistent with all other commands (`install`, `update`, `init`, `locale-apply`, etc.). Running `npx aioson setup:context . --defaults --lang=pt-BR` now correctly installs the pt-BR locale agents.
+
+## [0.1.21] - 2026-03-03
+### Changed
+- `@ux-ui` Step 0 (base + en/pt-BR/es/fr locales): added **Option C — Default / Skip** to the visual style question. When chosen, the agent skips style confirmation and goes directly to Step 1 using `interface-design.md` as the sole design authority, letting domain exploration drive the visual direction organically. Accepted inputs: C / skip / pular / saltar / passer / padrão / default.
+
+## [0.1.20] - 2026-03-03
+### Added
+- `template/.aioson/skills/static/react-motion-patterns.md`: new skill with React/Next.js equivalents of the wow effects from `static-html-patterns.md`. Covers 10 patterns: animated mesh background, animated gradient text, scroll reveal (Framer Motion + Intersection Observer fallback), 3D card tilt hook, hero staggered entrance, infinite logo marquee, scroll progress bar, glassmorphism card, floating orbs, and page transition. All patterns include `prefers-reduced-motion` fallback and touch-device guards.
+- Laravel installation links to `skills/dynamic/laravel-docs.md`: three-path table (Herd/standard, Sail/Docker, Jetstream+Livewire) with specific URLs, detection signals per stack, and Livewire v3 MCP fetch example.
+
+### Changed
+- `template/.aioson/skills/static/laravel-conventions.md`: added project folder structure tree, naming convention table (13 artefacts, singular/plural rules), and Livewire component section with `#[Computed]`, `wire:model.live`, classic controller variant, and coexistence rule.
+- `@dev` agent (base + en/pt-BR/es/fr locales): added **Motion and animation** section — when `framework=React|Next.js`, agent now reads `react-motion-patterns.md` before implementing animations. Framer Motion as primary, CSS `@keyframes` as fallback. Hard rule against heavy motion in admin/CRUD interfaces.
+- `@dev` agent (base + en/pt-BR/es/fr locales): added Laravel project structure layout block and singular/plural naming rules to the Laravel conventions section.
+
+## [0.1.19] - 2026-03-03
+### Added
+- `src/commands/test-agents.js`: new `test:agents` CLI command — 99 structural checks across all 8 base agents and all locale files (32 base checks + 64 locale checks + 3 critical skill checks). Validates: file existence, `## Mission` section, `## Hard constraints` section, `.md-only` context rule, and language instruction within first 15 lines of each locale agent.
+- `src/commands/locale-diff.js`: new `locale:diff` CLI command — compares `##` heading structure of base agent files vs locale translations, normalizes headings (strips accents for cross-language comparison), and reports sections present in base but missing in locale. Supports `--lang` filter and single-agent argument.
+- `template/.aioson/skills/static/django-patterns.md`: comprehensive Django conventions skill (342 lines) — project structure, `TimestampMixin`, service layer, DRF serializers, URL namespacing, settings split (base/dev/prod), Admin registration, `pytest-django` test patterns, N+1 prevention rules.
+- `template/.aioson/skills/static/fastapi-patterns.md`: comprehensive FastAPI conventions skill (344 lines) — async project structure, `pydantic-settings`, SQLAlchemy async engine, Pydantic input/output schema separation, service layer, JWT via `HTTPBearer`, Alembic migrations, `pytest + httpx AsyncClient` patterns.
+- `template/.aioson/skills/dynamic/README.md`: explains static vs dynamic skills pattern and lists all 6 dynamic skill files with their source URLs.
+
+### Changed
+- `@dev` agent (base + en/pt-BR/es/fr locales): `architecture.md` and `discovery.md` now marked as `*(SMALL/MEDIUM only)*` in Required input. Added explicit MICRO callout — only `project.context.md` is guaranteed for MICRO projects.
+- `@orchestrator` base agent: renamed `## Rules` → `## Hard constraints` for consistency with all other agents.
+- `@dev`, `@qa`, `@orchestrator` base agents: added `.md-only` context rule (was missing — `test:agents` detected the gap).
+- Locale `en` agents (analyst, architect, pm, ux-ui, dev, qa, orchestrator): added ABSOLUTE INSTRUCTION language marker at top (was missing in en locale — `test:agents` detected the gap).
+- All 6 dynamic skill stubs expanded with concrete source URLs, focus areas, and MCP tool fetch examples to replace the minimal placeholder content.
+
+## [0.1.18] - 2026-03-03
+### Fixed
+- Language bug: locale agent files (pt-BR, es, fr) were responding in English even after `aioson install --lang=<locale>` because the language rule was at the bottom of each file. LLMs process top-to-bottom and would default to English before reading the rule.
+- Added `> **⚠ ABSOLUTE INSTRUCTION — LANGUAGE**` blockquote as the **first content after the title** in all 8 agents across all 3 locales (pt-BR, es, fr) + en locale (24 files total). Instruction explicitly forbids English and names the target language before any mission, sequence, or detection logic runs.
+- `@ux-ui` output contract: HTML deliverable was being saved to `.aioson/context/landing-preview.html` instead of the project root. Fixed to `index.html` in the project root across all 5 ux-ui files (base + en/pt-BR/es/fr locales).
+
+## [0.1.17] - 2026-03-03
+### Added
+- `static-html-patterns.md` Section 0 — **Hero Law**: explicit rule that the hero MUST be full-viewport animated background + ONE headline + TWO buttons. Cards in the hero are forbidden.
+- `static-html-patterns.md` Section 2a-extra — **Mandatory Wow Techniques** for Bold & Cinematic (three required, not optional):
+  - Animated mesh background (`@keyframes meshDrift 20s`) — static gradients replaced
+  - Animated gradient text (`@keyframes textGradient 8s`) on headline `<em>` key phrase
+  - 3D card tilt on hover (`perspective(700px) rotateX/rotateY` on `mousemove`, skipped on touch + reduced motion)
+- `@setup` agent (base + all 4 locales) — **Step 3: Next agent guidance** appended to Post-setup action. Agent now closes by explicitly naming the next `@agent` based on `project_type` + `classification`, using the exact `@name` format so AI clients (Codex, Claude Code, Gemini) can trigger it.
+- `@setup` pt-BR/es/fr/en locales — spec.md skip hint for `project_type=site` + MICRO classification.
+
+### Changed
+- `@ux-ui` Step 0 (base + en/pt-BR/es/fr locales) — upgraded from "ask one question" to **HARD STOP blocking gate**: agent must not read files, write HTML/CSS, or proceed to Step 1 until user answers the visual style question.
+- `@ux-ui` Landing page mode (base + all locales) — added **Hero Law** constraint and **Mandatory Wow Techniques** section explicitly referencing the three required animations.
+- `@ux-ui` es/fr locales — **full rewrite** to match the updated base agent: added Step 0 (visual style intake), landing page mode, hero law, mandatory wow techniques, full output contract.
+
+## [0.1.16] - 2026-03-03
+### Added
+- `template/.aioson/skills/static/static-html-patterns.md`: new **Section 14 — Premium Template Patterns (Aigocy-style)** with 10 production patterns extracted from a real AI-agency landing page (ThemeForest #61450410):
+  - **14a. effectFade animations**: `fadeUp` and `fadeRotateX` (3D perspective entrance) with `data-delay` stagger pattern for GSAP
+  - **14b. Infinite logo marquee**: CSS-only `@keyframes infiniteSlide` with auto-clone JS and hover-pause for accessibility
+  - **14c. SVG animated paths**: SMIL `<animateMotion>` hub-and-spoke diagram connecting icons to a center product image
+  - **14d. Scroll-to-top with circular progress**: CSS `stroke-dashoffset` progress ring updated by scroll JS
+  - **14e. Split Swiper**: synchronized text + image sliders with `effect: 'fade'` and `slideTo()` binding
+  - **14f. Swiper progress bar navigation**: thin animated fill bar replacing pagination dots for portfolio sliders
+  - **14g. box-white / box-black section alternation**: CSS pseudo-element radial glow replacing decorative PNG images
+  - **14h. Accordion FAQ**: native `<details>`/`<summary>` version with `rotate(45deg)` icon transition
+  - **14i. Footer with watermark background logo**: faded brand name in `position: absolute` behind 3-column dense footer
+  - **14j. Canvas cursor trail**: fading dot trail on `mousemove`, skipped on touch devices and `prefers-reduced-motion`
+- Section 13 pre-delivery checklist: added "No placeholder text remains" item
+
+### Changed
+- `@setup` agent output template (`aioson_version`) bumped to `0.1.16` across all locales (en, pt-BR, es, fr) and base agent file
+
+## [0.1.12] - 2026-03-02
+### Added
+- New static skill: `template/.aioson/skills/static/interface-design.md` — a comprehensive UI/UX craft guide derived from the interface-design project, covering: Intent-First Framework (3 mandatory questions before any layout), Domain Exploration (4 required outputs: domain concepts, color world, signature element, defaults to avoid), 6 design directions with full token specs (Precision & Density, Warmth & Approachability, Data & Analysis, Editorial, Commerce, Minimal & Calm), complete token architecture (foreground/background/border/brand/semantic roles), depth strategy (commit to ONE), component state matrix, 4 quality checks (swap/squint/signature/token tests), self-critique process (composition → craft → content → structure).
+- Portuguese documentation at `docs/pt/`:
+  - `README.md`: index linking all guides.
+  - `inicio-rapido.md`: quick start with install commands, classification scoring, and 3-command setup.
+  - `agentes.md`: per-agent reference with when-to-use, activation command, what it delivers, and concrete examples.
+  - `cenarios.md`: 4 complete worked examples — MICRO (landing page), SMALL (Laravel clinic API with @analyst output, @architect folder structure, @dev code), MEDIUM (Next.js SaaS with 3-lane parallel orchestration), MEDIUM dApp (Ethereum NFT marketplace with Solidity contract example).
+  - `web3.md`: Portuguese Web3 guide covering setup flags for Ethereum/Solana/Cardano, monorepo structure, per-agent Web3 conventions, and skill reference.
+
+### Changed
+- `@ux-ui` agent (base + all 4 locales: en/pt-BR/es/fr): mandatory pre-work now references `interface-design.md` skill, adds Intent-First step, Domain Exploration step (4 required outputs), single design direction declaration, 4 quality checks (swap/squint/signature/token tests), and self-critique gate before delivery. Output contract expanded with focal point, reading order, full state matrix, and handoff notes for signature visual moves.
+- `workflow:plan` sequences corrected: `@ux-ui` now included in SMALL (`setup → analyst → architect → ux-ui → dev → qa`) and MEDIUM (`setup → analyst → architect → ux-ui → pm → orchestrator → dev → qa`), matching `config.md` and agent documentation.
+- 13 static skills expanded from 5–7 stub lines to 200–337 lines of production-grade, code-first references:
+  - `laravel-conventions`: controllers as orchestrators, Form Requests, Actions, Policies, Events+queued Listeners, Jobs, API Resources, N+1 prevention, Model conventions, Migrations, Pest tests.
+  - `tall-stack-patterns`: Livewire lifecycle + real-time validation, inter-component events, lazy loading, Alpine.js scope rules, Tailwind design system discipline, full Flux UI examples (button/modal/table/dropdown/sidebar).
+  - `filament-patterns`: Resource structure (form/table/filters/actions), custom Pages, Stats Widgets, Relation Managers, policy enforcement, advanced form fields (repeater, conditional, file upload), business logic delegation.
+  - `flux-ui-components`: full component inventory, buttons (all variants + loading states), form field group, modal with programmatic control, dropdown menus, table with sort + empty state, badges, sidebar navigation.
+  - `jetstream-setup`: Inertia vs Blade decision matrix, Teams with roles/permissions, API tokens, profile customization, 2FA, password confirmation middleware, post-install checklist.
+  - `git-conventions`: full Conventional Commits spec with type table + examples, Git Flow vs GitHub Flow, branch naming, PR template, tagging, protected history rules.
+  - `nextjs-patterns`: Server vs Client mental model, App Router structure, async Server Components, Server Actions with Zod, Client Components (when/why), Route Handlers for webhooks only, metadata/SEO, loading.tsx + error.tsx.
+  - `node-express-patterns`: layered architecture (routes/controllers/services/repositories), Zod validation middleware, typed auth + role checking, AppError class hierarchy, centralized error handler, rate limiting, graceful shutdown.
+  - `node-typescript-patterns`: strict tsconfig baseline, Zod at all external boundaries, env validation at startup, branded domain types for IDs, const-object enums, repository interface pattern, explicit return types, asyncHandler.
+  - `rails-conventions`: Service Objects with Result type, model scopes/enums/validations, Active Record N+1 prevention, async Jobs with retry, Mailer patterns, serializers, Pundit authorization, RSpec request + unit specs.
+  - `web3-ethereum-patterns`: CEI pattern, ReentrancyGuard, pull over push, AccessControl roles, gas optimization (struct packing, custom errors), Hardhat test patterns with loadFixture, wagmi v2 frontend integration, deployment scripts.
+  - `web3-solana-patterns`: eUTxO/account model, Anchor program structure, account data with LEN, PDA seeds + bump storage, CPI signing, Anchor constraints, compute budget management, full Anchor test suite.
+  - `web3-cardano-patterns`: eUTxO mental model, Aiken project structure, datum/redeemer type design, spending validator, minting policy, Aiken tests, off-chain with Lucid, datum versioning strategy, deployment checklist.
+  - `web3-security-checklist`: 7 critical vulnerabilities with code examples (reentrancy, access control, integer overflow, oracle manipulation, flash loans, front-running, signature replay), pre-deployment checklist (static analysis/fuzzing/invariant tests/multisig/timelock), emergency response protocol.
+
+### Fixed
+- `workflow:plan` was silently omitting `@ux-ui` from SMALL and MEDIUM sequences despite `config.md` and all agent documentation specifying it as a required step. Fixed in `src/commands/workflow-plan.js` and updated `tests/workflow-plan.test.js`.
+
+## [0.1.11] - 2026-03-02
+### Added
+- Agent prompt enrichment across all 8 agents:
+  - `@analyst`: 6 concrete Phase 1 discovery questions, entity deep-dive example (scheduling system), field-level table format for Phase 3, `Visual references` and `Risks identified` output sections, responsibility boundary note.
+  - `@architect`: concrete folder/module structure trees for MICRO/SMALL/MEDIUM across Laravel (TALL), Node/Express, Next.js (App Router), and dApp (Hardhat/Foundry/Anchor) stacks.
+  - `@pm`: explicit 2-page golden rule with cut-ruthlessly instruction, when-to-use guidance (SMALL/MEDIUM only, skip MICRO), exact `prd.md` section template.
+  - `@dev`: Laravel ALWAYS/NEVER convention list (Form Requests, Actions, Policies, Events+Listeners, Jobs, Resources, N+1 prevention), UI/UX conventions, Web3 guards for dApp projects, semantic commit format with examples, responsibility boundary note.
+  - `@orchestrator`: MEDIUM-only activation condition with early exit, 4-step orchestration process, dependency graph example, parallel vs sequential classification rules, `agent-N.status.md` and `shared-decisions.md` status file protocol.
+  - `@setup`: explicit `framework_installed` contract semantics (true/false downstream behavior), monorepo detection guidance for mixed Web3 + backend repos.
+  - `@ux-ui` and `@qa`: no structural changes (already complete).
+- All 8 locale packs (`en`, `pt-BR`, `es`, `fr`) synchronized with enriched agent content.
+- `isMonorepoDetection()` in `src/detector.js`: returns `true` when a Web3 framework and a backend or frontend framework coexist in the same directory.
+- Monorepo detection note propagated to `setup:context` output (localized via `note_monorepo` key in all 4 i18n message files).
+- `note_monorepo` i18n key added to `en`, `pt-BR`, `es`, and `fr` message dictionaries.
+- 4 new tests for `isMonorepoDetection` in `tests/detector.test.js`.
+
+### Changed
+- `setup:context` now prepends a localized monorepo warning note when Web3 and application framework signals coexist in the project directory.
+- `template/.aioson/config.md` context contract updated with explicit `framework_installed` semantics.
+- `aioson_version` example in `@setup` output template corrected from `0.1.8` to `0.1.10`.
+- `tests/agent-contracts.test.js` updated to reflect new `@dev` section names (`Laravel conventions`, `Responsibility boundary`).
+
+### Added
+- Full `pt-BR` CLI dictionary at `src/i18n/messages/pt-BR.js`.
+- Localized agent prompt packs for:
+  - `es` at `template/.aioson/locales/es/agents/*.md`
+  - `fr` at `template/.aioson/locales/fr/agents/*.md`
+- New `@ux-ui` agent contract and template set:
+  - `.aioson/agents/ux-ui.md`
+  - `.aioson/locales/{en,pt-BR,es,fr}/agents/ux-ui.md`
+  - `.gemini/commands/aios-ux-ui.toml`
+- Legacy framework detection support:
+  - `CodeIgniter 3`
+  - `CodeIgniter 4`
+- New package installation validation command:
+  - `aioson test:package [source-path] [--keep] [--dry-run] [--json]`
+- New workflow planning command:
+  - `aioson workflow:plan [path] [--classification=MICRO|SMALL|MEDIUM] [--json]`
+- `mcp:init` tool presets generation under `.aioson/mcp/presets/`:
+  - `claude.json`
+  - `codex.json`
+  - `gemini.json`
+  - `opencode.json`
+- New MCP validation command:
+  - `aioson mcp:doctor [path] [--strict-env] [--json]`
+- New parallel orchestration bootstrap command:
+  - `aioson parallel:init [path] [--workers=2..6] [--force] [--dry-run] [--json]`
+- New parallel diagnosis/remediation command:
+  - `aioson parallel:doctor [path] [--workers=2..6] [--fix] [--force] [--dry-run] [--json]`
+- New parallel scope assignment command:
+  - `aioson parallel:assign [path] [--source=auto|prd|architecture|discovery|<file>] [--workers=2..6] [--force] [--dry-run] [--json]`
+- New parallel consolidated status command:
+  - `aioson parallel:status [path] [--json]`
+- Optional `mcp:init` tool filter:
+  - `--tool=claude|codex|gemini|opencode`
+- `test:smoke` mixed monorepo profile:
+  - `--profile=mixed` for combined Web2 + Web3 workspace validation
+- `test:smoke` parallel orchestration profile:
+  - `--profile=parallel` to validate `parallel:init/assign/status/doctor` in one flow
+- Formal JSON schema documentation for automation:
+  - `docs/en/json-schemas.md`
+  - `docs/en/schemas/index.json`
+  - `docs/en/schemas/*.schema.json` for each JSON command contract
+  - includes `docs/en/schemas/mcp-doctor.schema.json`
+  - includes `docs/en/schemas/parallel-init.schema.json`
+  - includes `docs/en/schemas/parallel-doctor.schema.json`
+  - includes `docs/en/schemas/parallel-assign.schema.json`
+  - includes `docs/en/schemas/parallel-status.schema.json`
+
+### Changed
+- Version resolution is now centralized via `src/version.js` and shared across `info`, `setup:context`, and installer metadata generation.
+- Removed remaining hardcoded `0.1.8` fallback in `setup:context` version assignment.
+- Locale resolution now supports canonical fallback for Portuguese:
+  - `pt-BR` -> `pt-br`
+  - `pt_br` -> `pt-br`
+  - `pt` -> `pt-br`
+- Locale resolution now also supports regional fallback for:
+  - `es-*` -> `es`
+  - `fr-*` -> `fr`
+- Agent locale resolution now supports:
+  - `es-*` -> `es`
+  - `fr-*` -> `fr`
+- `update` now reapplies active agent prompts from `conversation_language` in context, preventing locale reset after template refresh.
+- `mcp:init` now writes both shared plan and tool-specific preset templates.
+- `mcp:init` Context7/Database presets now default to remote-endpoint bridge templates (`mcp-remote`) with URL env vars instead of generic command placeholders.
+- `mcp:init` invalid `--tool` errors are now localized via i18n dictionaries.
+- `mcp:doctor` human-readable check messages and summary are now fully localized.
+- `workflow:plan` advisory notes are now localized from i18n keys instead of hardcoded English.
+- `setup:context` onboarding notes are now localized (including beginner recommendation notes and stack-option notes).
+- `test:package` failure messages are now localized via i18n dictionaries.
+- `parallel:doctor` check/hint messages are now fully localized (including severity labels).
+- Internal validation failures inside `test:smoke` are now localized instead of hardcoded English.
+- `parallel:status` human-readable status rows and lane summaries are now localized (including status labels).
+- `mcp:doctor` severity prefixes are now localized per locale (`OK/WARN/FAIL` equivalents).
+- `test:package` now localizes fallback command failure detail when stderr/stdout are empty.
+- `parallel:assign` lane scope summary lines are now localized in human-readable output.
+- Parallel commands now localize fallback `unknown` classification labels in human-readable errors.
+- `mcp:init` now avoids hardcoded fallback text for invalid `--tool` and uses i18n-backed messaging.
+- `mcp:init` now localizes generated server `reason` fields and preset `notes` content via i18n.
+- Removed remaining hardcoded `unknown` fallbacks in `parallel:doctor` check messages and `context:validate` parse-reason output.
+- `context:validate` now localizes known frontmatter parse reason codes into human-readable locale messages.
+- `mcp:doctor` now localizes context frontmatter parse reason codes using the same i18n mapping as `context:validate`.
+- Standardized localized line formatting for `agents`, `locale:apply`, `workflow:plan`, and `parallel:init` human-readable listings.
+- Localized diagnostic line wrappers for `mcp:doctor` and `parallel:doctor` checks/hints, plus `context:validate` issue list rows.
+- Localized diagnostic/action wrapper lines in `doctor` command output (checks, hints, fix actions, and detail lines).
+- `cli` help and unknown-command wrapper lines are now localized via i18n keys instead of inline formatting.
+- `init` and `install` now print explicit multi-IDE onboarding hints (`agents` + `agent:prompt setup`) with optional `--tool=codex|claude|gemini|opencode` guidance.
+- Agent flow now supports explicit UI/UX handoff:
+  - `@architect` may hand off key screens/component constraints to `@ux-ui`
+  - `@dev` now consumes `.aioson/context/ui-spec.md` when present
+  - `SMALL`/`MEDIUM` default sequence includes `@ux-ui`
+- `doctor` now enforces multi-IDE gateway contracts (Claude/Codex/Gemini/OpenCode) and validates required Gemini/OpenCode files.
+- `doctor` now also validates each Gemini command file (`.gemini/commands/aios-*.toml`) maps to the expected shared agent instruction file.
+- `doctor --fix` now restores broken gateway contract files (Claude/Codex/Gemini/OpenCode + Gemini command mappings) from template in safe mode.
+- `update` now supports `--lang=en|pt-BR|es|fr` (and `--language`) to force localized agent-pack sync during update, including dry-run planning.
+- `--json` output mode now supports `init`, `install`, and `update` with clean machine-readable payloads (no mixed human logs).
+- `--json` output mode now also supports `agents`, `agent:prompt`, `locale:apply`, `setup:context`, and `i18n:add`.
+- JSON payloads for `init/install/update/agents/agent:prompt/locale:apply/setup:context/i18n:add` now include stable `ok` and command context fields.
+- JSON schema catalog expanded with formal contracts for:
+  - `init`
+  - `install`
+  - `update`
+  - `agents`
+  - `agent:prompt`
+  - `locale:apply`
+  - `setup:context`
+  - `i18n:add`
+- `init` and `install` now support `--lang=en|pt-BR|es|fr` to auto-apply localized agent packs during bootstrap.
+- Setup templates now default `aioson_version` to `0.1.8`.
+- `ui-ux-modern` static skill was expanded to a production-ready checklist with token, state, accessibility, responsive, and handoff guidance.
+- Developer onboarding now handles `Other` backend/frontend choices as true free-text custom values (legacy/custom stacks).
+- Expanded automated coverage:
+  - `tests/i18n-cli.test.js`
+  - `tests/init-install-guidance.test.js`
+  - `tests/prompt-tool.test.js`
+  - `tests/init-install-guidance.test.js` now validates locale pack auto-apply on `init/install --lang`
+  - `tests/version.test.js`
+  - `tests/update.test.js` now covers `--lang` override and dry-run locale sync planning
+  - `tests/json-output.test.js` now covers `init/install/update --json`
+  - `tests/json-output.test.js` now covers `agents/agent:prompt/locale:apply/setup:context/i18n:add --json`
+  - `tests/json-schema-files.test.js` now validates expanded schema catalog
+  - `tests/locales.test.js`
+  - `tests/agent-contracts.test.js`
+  - `tests/smoke.test.js` now verifies active agent prompt locale application for `--lang=es` and `--lang=fr`
+  - `tests/mcp-doctor.test.js`
+  - `tests/mcp-init.test.js` now covers invalid `--tool` handling
+  - `tests/setup-context.test.js` now validates localized onboarding notes
+  - `tests/parallel-doctor.test.js` now validates localized check messages
+  - `tests/parallel-init.test.js`
+  - `tests/parallel-doctor.test.js`
+  - `tests/parallel-assign.test.js`
+  - `tests/parallel-status.test.js`
+  - `tests/json-output.test.js` now covers `mcp:doctor --json`
+  - `tests/package-test.test.js`
+  - `tests/workflow-plan.test.js`
+  - extended `tests/mcp-init.test.js` and `tests/json-output.test.js`
+  - smoke JSON e2e checks for `--locale=es` and `--locale=fr`
+  - smoke coverage for `--profile=parallel`
+
+## [0.1.7] - 2026-03-01
+### Added
+- JSON output mode (`--json`) for:
+  - `aioson info`
+  - `aioson doctor`
+  - `aioson context:validate`
+  - `aioson test:smoke`
+- New JSON output test suite: `tests/json-output.test.js`.
+
+### Changed
+- CLI now returns structured JSON errors for unknown commands and runtime failures when `--json` is enabled.
+- `setup:context` and setup templates now default `aioson_version` to `0.1.7`.
+
+## [0.1.6] - 2026-03-01
+### Added
+- `test:smoke` now supports chain-specific Web3 profiles:
+  - `--web3=ethereum`
+  - `--web3=solana`
+  - `--web3=cardano`
+- Web3 smoke workflow now verifies:
+  - framework detection per chain profile
+  - `project.context.md` dApp/Web3 frontmatter consistency.
+
+### Changed
+- CLI help and docs updated for `test:smoke --web3`.
+- `setup:context` and setup templates now default `aioson_version` to `0.1.6`.
+
+## [0.1.5] - 2026-03-01
+### Added
+- Web3 framework detection:
+  - Ethereum: `Hardhat`, `Foundry`, `Truffle`
+  - Solana: `Anchor`, `Solana Web3`
+  - Cardano: `Cardano` (Aiken/Cardano SDK signals)
+- New Web3 skill templates:
+  - static: `web3-ethereum-patterns`, `web3-solana-patterns`, `web3-cardano-patterns`, `web3-security-checklist`, `node-typescript-patterns`
+  - dynamic: `ethereum-docs`, `solana-docs`, `cardano-docs`
+- New documentation page: `docs/en/web3.md`.
+
+### Changed
+- `project_type` now accepts `dapp`.
+- `setup:context` now supports Web3 context fields (`web3_enabled`, `web3_networks`, `contract_framework`, `wallet_provider`, `indexer`, `rpc_provider`).
+- `setup:context` and setup templates now default `aioson_version` to `0.1.5`.
+
+## [0.1.4] - 2026-03-01
+### Added
+- New command:
+  - `aioson test:smoke [workspace-path] [--lang=en|pt-BR] [--keep]`
+- New smoke test suite: `tests/smoke.test.js`.
+
+### Changed
+- `doctor` now supports safe remediation mode:
+  - `aioson doctor --fix`
+  - `aioson doctor --fix --dry-run`
+- `setup:context` and setup templates now default `aioson_version` to `0.1.4`.
+
+## [0.1.3] - 2026-03-01
+### Added
+- Localized agent prompt packs:
+  - `.aioson/locales/en/agents/*.md`
+  - `.aioson/locales/pt-BR/agents/*.md`
+- New command:
+  - `aioson locale:apply [path] [--lang=en|pt-BR] [--dry-run]`
+- Agent path resolution now supports locale-aware prompts with fallback to active agent paths.
+- New locale test suite: `tests/locales.test.js`.
+
+### Changed
+- `setup:context` now applies localized agent prompts based on `conversation_language`.
+- Documentation updated for locale pack workflow.
+
+## [0.1.2] - 2026-03-01
+### Added
+- New commands:
+  - `aioson agents`
+  - `aioson agent:prompt <agent> [--tool=...]`
+  - `aioson context:validate [path]`
+  - `aioson setup:context [path]` (interactive or defaults mode)
+- New modules:
+  - `src/context.js` for frontmatter parsing and context contract validation
+  - `src/agents.js` for agent metadata and prompt generation
+  - `src/context-writer.js` for context rendering and classification scoring
+- New test suites:
+  - `tests/context.test.js`
+  - `tests/agents.test.js`
+  - `tests/context-writer.test.js`
+
+### Changed
+- `doctor` now validates `project.context.md` frontmatter and required fields, including `conversation_language` format.
+
+## [0.1.1] - 2026-03-01
+### Changed
+- Converted remaining template files to English (`.aioson` agents/config and Gemini command descriptions).
+- Hardened `@setup` instructions to prevent silent defaults for `project_type`, `profile`, `classification`, and language.
+- Added explicit `conversation_language` to context contract so agent interaction language can be enforced.
+
+## [0.1.0] - 2026-03-01
+### Added
+- Initial CLI commands: init, install, update, info, doctor
+- Multi-IDE template gateways (Claude, Codex, Gemini, OpenCode)
+- Framework detector and installer/updater core
+- i18n message system with English default
+- Automated tests for detector, installer, doctor, i18n
+- `i18n:add <locale>` command to scaffold new locale dictionaries
+- GitHub Actions CI and tag-based npm release workflows
+
+### Changed
+- Project-facing content standardized to English
+- CLI i18n upgraded with dynamic locale loading and fallback behavior