# Changelog

All notable changes to GameDesignOS will be recorded here.

## [Unreleased]

### Added

- Added the candidate UL (Uncertainty Ladder) control layer to `paranoia-ai-system-evolver`: canonical `UL-L0`–`UL-L5` rungs, `ul_state`, a machine-readable schema and example, optional workflow-run reference, bottleneck attribution, progressive scaffold removal, transfer and negative-transfer checks, routing signals, validators, and runtime regressions.
- Added generated wheel snapshots of canonical contracts and workspace templates so installed runtimes can operate without a source checkout, including an sdist-to-wheel rebuild gate.
- Added Agent Skills portability metadata, 7/7 five-gate behavior suites without dropping existing domain regressions, cross-platform CI, an installed-wheel smoke test, and the public-synthetic Golden Lighthouse path.
- Added `SECURITY.md` and Dependabot configuration.
- Added a text-free GitHub hero visual that shows fragmented AI game-design inputs resolving into evidence, experiments, decisions, and durable project memory.
- Added a dated GitHub README benchmark covering Browser Use, OpenHands, MetaGPT, AutoGen, CrewAI, and LangGraph with explicit transfer and non-transfer boundaries.

### Changed

- Changed the development runtime to `1.3.0.dev0` while keeping workspace schema `1.0.0` compatible.
- Made `contracts/router.yaml` the only editable runtime routing source; Python now compiles deterministic rules from the contract.
- Kept UL optional and `shadow` by default: it controls uncertainty exposure between VOI selection and OODA execution, but does not replace domain skills, migrate existing workspaces, or expand Human Gate authority.
- Centralized the package/runtime candidate version in `gamedesignos/_version.py`, removed the hard-coded version from the CI wheel-install command, and declared Python 3.11/3.12/3.13 package support.
- Upgraded official GitHub Actions to Node 24-compatible `actions/checkout@v7` and `actions/setup-python@v7`, added read-only workflow permissions, concurrency cancellation, and a bounded job timeout.
- Reconciled the repository architecture, GitHub About source, and release-state contract with the current public repository: 13 remote tags, a verified Latest GitHub Release for `v1.2.0`, no PyPI package, and `v1.3.0.dev0` still candidate.
- Added the two packaged static behavior-eval fixture suites to CI.
- Added trustworthiness regressions for README schema counts and runtime/schema decision-type parity.
- Refreshed product architecture, roadmap, usage, and v1.2 release-status documentation to match the current tagged runtime.
- Reworked the README first screen around one outcome promise, four primary links, a visible decision flow, and a 30-second Quick Start; refreshed GitHub About copy and topics to match.

### Fixed

- Made CLI stdout/stderr, installed-wheel smoke capture, and every nested smoke subprocess use an explicit UTF-8 contract so Chinese output does not crash or become locale-dependent on redirected Windows streams.
- Restored isolated sdist-to-wheel rebuilding so clean Python 3.12/3.13 environments install the declared `setuptools.build_meta` backend instead of relying on ambient build tools.
- Completed the static v1 workspace template with its declared `00-inbox/` guide and added the file to package smoke coverage.
- Aligned v1 Assumption, Evidence, Experiment, and Workflow runtime records with their canonical schemas and added a full-cycle schema regression.
- Corrected the no-Decision next-action hint to use `gamedesignos decision new ...`.
- Made `gamedesignos ask` route-only unless the user supplies `--destination` / `--workspace`; routing confidence no longer acts as authorization to create files, and explicit long-lived setup remains available through `start`.
- Converted filesystem failures at the CLI boundary into concise validation errors instead of exposing raw tracebacks.
- Added negative routing regressions for project-review and high-confidence project requests without an explicit destination.

### Security

- Removed skill-root escape references and kept publishing, durable rule promotion, real-account actions, and irreversible commitments behind Human Gate.

## [v1.2.0] - 2026-07-09

### Added

- Added Intent Work Order support across `paranoia-ai-system-evolver`, including schema, templates, playbook guidance, routing, README prompts, behavior evals, and validation checks.
- Added `workflow-run.governance` so workflow runs can preserve intent, Decision/VOI, RJR authority, Paranoia review, Human Gate, rollback, candidate learning, and retrospective references.
- Added Project Workflow Governance guidance and `workflow_governance_review` templates so `paranoia-ai-system-evolver` participates as a cross-workflow governance layer without replacing domain skills.
- Added Paranoia Checkpoints to all documented end-to-end workflows.

### Changed

- Changed the runtime package version to `1.2.0` while keeping the Project-Ready workspace schema at `1.0.0`.
- Updated README trio, runtime docs, package metadata, v1 workspace manifest, project-workspace contract, router rules, runtime workflow creation, tests, and validators for v1.2.
- Updated release status to account for missing historical local tags and the v1.2 release note.

### Safety

- Kept workflow governance defaulted to `shadow` so it records drift and boundary findings without blocking low-risk domain work.
- Kept `paranoia-ai-system-evolver` as a governance layer rather than a mega-agent that owns concept, diagnosis, ED experiment, or proposal outputs.
- Tagged the reviewed v1.2 release commit; later schema-hardening and CLI fixes remain under `Unreleased` until the next release.

## [v1.1.0] - 2026-07-02

### Added

- Added RJR-AI as the explicit residual-judgment authority layer for AI engineering workflows: AI expands possibilities, Workflow compresses chaos, Eval provides feedback, permission gates prevent overreach, knowledge bases preserve memory, and humans retain high-coupling, low-reversibility, under-evidenced decisions.
- Added RJR-AI coverage to `paranoia-ai-system-evolver`, including skill guidance, references, templates, eval cases, router wording, and validation checks.
- Added a v1.1 release note and GitHub-facing metadata guidance for the new project description and topic set.

### Changed

- Changed the runtime package version to `1.1.0` while keeping the Project-Ready workspace schema at `1.0.0`.
- Updated README trio, runtime docs, product vision, roadmap, GitHub About checklist, package metadata, and v1 workspace manifest to the v1.1 positioning.
- Updated natural-language routing so RJR-AI, residual judgment, permission-boundary, workflow/eval/memory, and authority-layer requests route to `paranoia-ai-system-evolver`.

### Safety

- Kept the installed runtime-copy boundary explicit: source-repo validation does not automatically mutate a user's installed skill directory.
- Kept Human Gate as the owner of residual judgment; RJR-AI organizes delegation and constraints but does not let automation accept irreversible or commitment-changing choices.
- Kept existing v1 workspaces compatible because this release does not change `schema_version`.

## [v1.0.0] - 2026-06-19

### Added

- Added Project-Ready v1 workspace structure: `01-decisions`, `02-assumptions`, `03-evidence`, `04-experiments`, `05-design-assets`, `06-workflows`, `07-learning`, and `.gamedesignos` runtime state.
- Added v1 core contracts: Decision, Assumption Registry, Evidence Ledger, Experiment Plan, Experiment Result, Learning Record, Gate Result, and Workflow Run.
- Added v1 CLI groups: `decision`, `assumption`, `evidence`, `experiment`, `gate`, `workflow`, `health`, `next`, and `graph`.
- Added `gamedesignos start` as a one-command onboarding path that creates or resumes a v1 workspace with the first Decision, Assumption, Experiment, VOI Gate, and Workflow Run.
- Added `gamedesignos ask` and freeform `gamedesignos "<one sentence>"` routing so cloned repositories can accept a natural-language request immediately.
- Added root `AGENTS.md` and Chinese usage docs that explain whole-project use and each standalone skill route.
- Added deterministic gates for VOI, Evidence, Scope, Experiment, Commitment, and Rollback.
- Added Decision Graph Mermaid export and graph inspection.
- Added a complete runtime behavior test for the chain: Decision -> Assumption -> Experiment Plan -> Evidence -> Result -> Review -> Assumption validation -> Decision accept -> Workflow validation.
- Added Chinese-first Project-Ready documentation for runtime, CLI commands, contracts, roadmap, v1 workspace template, and release notes.
- Added a formal v1.0 vector README hero with crisp English project name, core positioning, asset counts, and safety boundary: `assets/gamedesignos-v1-hero.svg`.

### Changed

- Changed the runtime package version to `1.0.0`.
- New workspaces now default to schema `1.0.0`; legacy schema `0.8.0` remains available through `--workspace-version 0.8.0`.
- Updated README entrypoints to present v1.0.0 as Project-Ready rather than a local prototype.
- Updated the README trio to use the formal v1.0 SVG hero instead of the v0.9 runtime hero.
- Updated repository validation to require the v1 contract and workspace-template surfaces.

### Safety

- Decision acceptance requires explicit `--by` and `--reason`.
- Commitment gate blocks high-impact decisions without rollback, untested high-risk assumptions, and unreviewed linked experiments.
- Runtime remains deterministic and local-only: no model calls, uploads, credentials, publishing, or automatic Human Gate decisions.
- Rollback remains file-scoped and does not require moving project roots, replacing `.git`, or mirror-syncing workspaces.

## [v0.9.0] - 2026-06-19

### Added

- Added the first executable `gamedesignos` local runtime with `init`, `status`, `voi`, `route`, `new`, `validate`, `pack`, and `doctor` commands.
- Added an installable Python package with both `gamedesignos` and `python -m gamedesignos` entrypoints.
- Added deterministic workspace initialization, draft asset creation, qualitative VOI assessment/review, route recommendation, structural validation, and review-safe packaging.
- Added runtime tests covering private and public-synthetic workspace flows, missing-upstream routing, VOI boundaries, validation failures, and pack filtering.
- Added a generated v0.9.0 runtime hero image for the README trio: `assets/gamedesignos-runtime-v09-hero.png`.

### Changed

- Decoupled workspace schema version `0.8.0` from runtime implementation version `0.9.0`.
- Updated the project-workspace schema to accept v0.8 and v0.9 runtime declarations.
- Updated runtime documentation from planned command contracts to executable CLI behavior.
- Updated the README trio to present v0.9.0 as an executable local runtime rather than only a Runtime Foundation.
- Preserved direct installation and invocation of all seven existing skill packages.

### Safety

- Kept model calls, credentials, uploads, automatic skill execution, and Human Gate decisions outside the runtime.
- Kept `--force` from overwriting existing GameDesignOS workspaces.
- Kept draft assets unaccepted by default and packed outputs limited by registered asset source status.

## [v0.8.0] - 2026-06-18

### Added

- Added the GameDesignOS Runtime Foundation: product vision, four-layer architecture, MVP boundary, and capability-gated roadmap.
- Added a copyable project workspace template with a manifest, lifecycle directories, design-asset index example, and Human Gate decision-log example.
- Added workspace-level schemas for project manifests, design-asset indexes, and decision logs.
- Added four workspace-aware workflow guides: idea-to-validation, media-to-diagnosis, weekly ED experiment, and evidence-to-proposal.
- Added planned local CLI command contracts without claiming that a complete CLI ships in this version.
- Added a complete VOI Decision Gate for Decision Object, current default action, decision boundary, signal-to-action mapping, EVPI/EVSI, information costs, smallest probes, and stop rules.
- Added `information-value-assessment` as a workspace contract and durable decision asset.
- Added VOI behavior evals for FOMO research, branch explosion, local negative evidence, irreversible changes, and high-structure/low-value output.

### Changed

- Repositioned the README trio around `Skill Kernel + Contract Layer + Project Workspace + Runtime Interface`.
- Reused the existing no-text hero background so the v0.7.0 title layer is not shown on a v0.8.0 page.
- Upgraded adapter guidance from direct skill loading to workspace-aware routing and asset writeback.
- Extended repository validation to protect runtime paths, workspace conventions, schemas, examples, and the VOI Decision Gate.
- Upgraded `paranoia-ai-system-evolver` so research, retrieval, tests, memory reads, and AI branches must justify how they can change action.
- Preserved direct installation and invocation of all seven existing skill packages.

### Safety

- Kept real project material in private workspaces or overlays outside the public repository.
- Kept Human Gates for project commitments and publication.
- Kept API keys, credentials, billing, and host permissions outside GameDesignOS.

## [v0.7.0] - 2026-06-18

### Added

- Added the `GameDesignOS by Paranoia` v7 README banner assets with a no-text background and deterministic local title/signature layer: `assets/gamedesignos-overview-banner-v7-background.png` and `assets/gamedesignos-overview-banner-v7.png`.
- Added the Dreamina-generated `assets/showcase-game-design-proposal-writer.png` so all seven public skills now have visible README showcase art.

### Changed

- Updated canonical public repository references to `DY-2026/GameDesignOS`, including README Star History, GitHub About metadata, showcase feedback links, and schema identifiers.
- Refined the public visual direction toward `GameDesignOS` as an AI-native game design operating system rather than a skill-list wrapper.
- Consolidated the `《生存33天》` Game Experience Analyzer report and the `《冒险家艾略特的千年奇谭》` ED proof path into the featured cases table so the 60-second demo stays a quick interaction example.
- Reorganized the README showcase grid into seven clickable module cards following the GameDesignOS product path: concept, evidence, proposal, ED iteration, workflow governance, translation, and source curation.
- Updated current adapter docs, skill sub-README parent links, and the ED experiment schema id from the old `ParanoiaSkills` name to `GameDesignOS` / `GameDesignOS by Paranoia`.
- Marked the older public article as a historical `ParanoiaSkills` draft so it does not conflict with the v0.7.0 identity.

## [v0.6.1] - 2026-06-10

### Fixed

- Re-added repository `.gitignore` as a required tracked file so CI validation can pass without repository-local exceptions.

- Documented the release housekeeping so the versioning sequence remains accurate after the hotfix.

## [v0.6.0] - 2026-06-09

### Added

- Added `game-design-proposal-writer` as the seventh public skill package for turning research, concept contracts, evidence notes, validation plans, production constraints, and business goals into reviewable proposals, pitch outlines, decision memos, and vertical-slice documents.
- Added proposal-writing references, templates, evals, negative cases, synthetic outputs, and public synthetic examples.
- Added `game-experience-density-optimizer` evidence-gate and scorecard surfaces for weekly ED experiment planning, schema output, hybrid-conflict review, instrumentation, and decision rules.
- Added a new GitHub landing banner asset for the seven-skill architecture: `assets/paranoia-skills-overview-banner-v6.png`.

### Changed

- Updated the README trio, showcase index, router contracts, and repository validation for the seven-skill architecture.
- Repositioned the public GitHub intro around a verifiable agent-skill operating system for game design: evidence diagnosis, concept contracts, proposal writing, ED experiments, and workflow governance.
- Upgraded `game-experience-density-optimizer` from a methods-heavy skill into an evidence-gated experiment compiler with explicit modes, templates, schemas, examples, and eval coverage.

### Safety

- Kept public examples synthetic or explicitly bounded by the repository contribution policy.
- Generated the new banner from a no-text background and overlaid exact project/version text locally to avoid model-generated typo artifacts.

## [v0.5.0] - 2026-06-06

### Changed

- Refreshed the public README introduction around cross-skill contracts, ED experiments, WOOP/VOI/OODA workflow governance, and eval-backed evolution.
- Added a new overview banner asset with project name, one-line positioning, and version marker for the GitHub landing page.
- Updated the GitHub About description for the new contracted workflow positioning.

### Added

- Added cross-skill routing and handoff language to the public introduction, including concept contracts, evidence indexes, issue cards, ED handoffs, and validation plans.
- Added WOOP Task Card positioning to the public workflow-evolution language.

## [v0.4.0] - 2026-06-04

### Added

- Added `game-experience-density-optimizer` for turning experience concentration, retention, pacing, feedback, embodiment, atmosphere, and cognitive-load problems into weekly ED experiments.
- Added ED templates, telemetry/dashboard specs, evals, a synthetic public example, and showcase imagery.

### Changed

- Updated the README trio, onboarding guide, showcase index, and repository validator for the six-skill architecture.
- Updated the overview/showcase layout so visual cards render four per row instead of squeezing six into one row.

## [v0.3.0] - 2026-06-03

### Added

- Added star-supported public content promise.
- Added try-it-in-10-minutes onboarding.
- Added showcase index.
- Added examples index.
- Added issue templates.
- Added v0.3.0 release draft.
- Added Star History chart to the README trio.
- Reworked the README opening so positioning and value come before specific examples.
- Added a generated overview banner with title and one-line positioning.
- Simplified repository layout docs into reader-facing layers instead of a full asset tree.

## [v0.2.0] - 2026-05-30

### Added

- Added `game-concept-architect` as a public base skill for turning one-line game ideas into verifiable concept blueprints.
- Added adapters, contracts, repository validation, and contribution boundaries for public-safe examples.

### Changed

- Reorganized the repository into skill packages, adapters, contracts, docs, releases, and validation scripts.
- Strengthened the public/private boundary for examples, assets, evals, and docs.

## [v0.1.0] - 2026-05-29

### Added

- Released the public base version of `game-experience-analyzer`.
- Added initial release notes, installation guidance, and validation examples.

[Unreleased]: https://github.com/DY-2026/GameDesignOS/compare/v1.2.0...HEAD
[v1.2.0]: ./releases/v1.2.0.md
[v1.1.0]: ./releases/v1.1.0.md
[v1.0.0]: ./releases/v1.0.0.md
[v0.9.0]: ./releases/v0.9.0.md
[v0.8.0]: ./releases/v0.8.0.md
[v0.7.0]: ./releases/v0.7.0.md
[v0.6.1]: ./releases/v0.6.1.md
[v0.6.0]: ./releases/v0.6.0.md
[v0.5.0]: ./releases/v0.5.0.md
[v0.4.0]: ./releases/v0.4.0.md
[v0.3.0]: ./releases/v0.3.0.md
[v0.2.0]: ./releases/v0.2.0.md
[v0.1.0]: ./releases/v0.1.0.md