devrites 3.2.27 → 4.0.0

package/CHANGELOG.md

@@ -2,6 +2,28 @@
 
 All notable changes to DevRites are documented here. The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and DevRites adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). Releases are generated automatically by [semantic-release](https://semantic-release.gitbook.io/) from Conventional Commits on `main`.
 
+## [4.0.0](https://github.com/ViktorsBaikers/DevRites/compare/v3.2.28...v4.0.0) (2026-08-03)
+
+### ⚠ BREAKING CHANGES
+
+* **repo:** use native host orchestration
+
+### Changed
+
+* **repo:** use native host orchestration ([0f871b2](https://github.com/ViktorsBaikers/DevRites/commit/0f871b2e2680290ed707006c693a91453619e079))
+
+### Fixed
+
+* **docs:** synchronize native orchestration contracts ([d80aaed](https://github.com/ViktorsBaikers/DevRites/commit/d80aaedf83d598b201060f51f0c58f570a7d676f))
+* **tests:** admit LF index fixture on Windows ([1d3cc3b](https://github.com/ViktorsBaikers/DevRites/commit/1d3cc3bc693f35ea0b1f4d2ce0eeb0f3ba3006a0))
+* **tests:** make Git fixtures portable on Windows ([287bcf6](https://github.com/ViktorsBaikers/DevRites/commit/287bcf65d15ef4a2a56913b358bee2ea22dff0b1))
+
+## [3.2.28](https://github.com/ViktorsBaikers/DevRites/compare/v3.2.27...v3.2.28) (2026-07-29)
+
+### Fixed
+
+* **agents:** bind durable V2 agent starts ([1da2cef](https://github.com/ViktorsBaikers/DevRites/commit/1da2cefd2a1e6b2ecb6a626224cc94fa0fc3662e))
+
 ## [3.2.27](https://github.com/ViktorsBaikers/DevRites/compare/v3.2.26...v3.2.27) (2026-07-29)
 
 ### Fixed
@@ -665,147 +687,3 @@ Not-tested: live model-backed Codex exec and custom-subagent smoke are opt-in.
 
 * **ci:** bypass commitlint on semantic-release commits, tidy README ([0cf52a3](https://github.com/ViktorsBaikers/DevRites/commit/0cf52a3f27216ab5edb8197ec22628d03f2e5e31))
 * **ci:** sync lockfile and reject multiline descriptions without PyYAML ([0efa85f](https://github.com/ViktorsBaikers/DevRites/commit/0efa85f052612b94926cf3382f88510059e5a8e8))
-
-# Changelog
-
-All notable changes to DevRites are documented here. The format is based on
-[Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and DevRites adheres
-to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
-
-## [Unreleased]
-
-### Added
-
-- **AFK & HITL run modes — full pause/resume contract.**
-  - `pack/.claude/skills/rite-resolve/` — new public skill (`SKILL.md`,
-    `scripts/resolve.sh`, `reference/answer-protocol.md`). Canonical writer
-    for `questions.md` `status: open → answered` and `state.md` `Awaiting
-    human` clearance. Three input shapes: single `<qid> "<answer>"`,
-    `--drop <qid>`, `--batch <file>`.
-  - `pack/.claude/rules/afk-hitl.md` — master contract: `.devrites/AFK`
-    sentinel format, four-gate taxonomy (`advisory` / `validating` /
-    `blocking` / `escalating`) with SLAs (`none` / `4h` / `15m` / `24h`),
-    irreversible-risk list (destructive migration, auth/authz boundary,
-    public API break, external contract, fs destruction, red tests/types/
-    lint) that always pauses regardless of `allow_gates`.
-  - `pack/.claude/skills/rite-define/reference/gates.md` — gate decision tree
-    + per-gate behavior matrix + AFK interaction table.
-  - `pack/.claude/skills/rite-build/reference/checkpoint-protocol.md` —
-    pre-action render contract + workspace mutations + `notify:` hook env
-    contract.
-  - `pack/.claude/skills/rite-build/reference/afk-discipline.md` — iteration
-    cap, fail-on-red, irreversible-risk list, notify-hook seam.
-  - `.devrites/AFK` sentinel (project-root, gitignore by default) toggles
-    session-level AFK mode. Optional YAML: `max_slices`, `notify`,
-    `allow_gates`. Empty file = AFK with safe defaults
-    (`allow_gates: [advisory]`).
-- **Schema extensions** (`pack/.claude/skills/rite-spec/reference/state-workspace.md`):
-  `state.md` gains `Run mode`, `Status`, `Slice mode`, and an `Awaiting
-  human` block when paused. `tasks.md` slice format extended with `Gate`,
-  `SLA`, `Checkpoint` (required when `Mode: HITL`). `questions.md` entries
-  carry `qid`, `gate`, `status`, `proposed`, `raised_at`, `answered_at`,
-  `answer`.
-- `/rite-build` workflow updated with:
-  - Step 0 — awaiting check (stop and route to `/rite-resolve` when
-    `Status: awaiting_human`).
-  - Step 2a — HITL pre-action checkpoint (renders gate, persists
-    `Awaiting human`, fires `notify:` hook, stops).
-  - Step 10 — fail-on-red (red tests/types/lint never marks the slice
-    `built`; raises a blocking question).
-  - Step 11 — sentinel decrement (`max_slices` ticks down per built slice;
-    0 forces HITL stop).
-- `/rite-status` surfaces `Run mode`, `Status`, the `Awaiting human` block,
-  and open questions broken down by gate (`blocking · validating · advisory
-  · escalating`). `scripts/load-state.sh` parses `.devrites/AFK` + tallies
-  questions by gate.
-- `devrites-doubt` AFK exception: findings at or below the slice's gate
-  ceiling downgrade to advisory entries in `questions.md`; irreversible-risk
-  findings always pause regardless of AFK config.
-- README "Modes — HITL & AFK" section + top-of-doc 2-mode callout + Contents
-  link + phase-table `RESUME` row pointing at `/rite-resolve`.
-- `docs/usage.md` — new sections "9) HITL gate — pre-code pause and resume"
-  and "10) AFK overnight run"; `.devrites/AFK` sentinel row in the workspace
-  table.
-- `docs/flow.md` — feature lifecycle diagram now shows the `Awaiting human`
-  ↔ `/rite-resolve` loop; workspace state-model ER diagram extended with
-  `AFK_SENTINEL`, `state.run_mode`/`status`/`awaiting_human`, and
-  `questions.qid`/`status`/`gate`; new section 9 "AFK & HITL state
-  machine".
-- `docs/architecture.md` — new "Run modes — HITL & AFK" subsection +
-  Surface bullet updated with `rite-resolve`.
-- `docs/command-map.md` — `/rite-resolve` row + updated `/rite-status` row
-  (reports run mode + status + question gate breakdown).
-- `.claude-plugin/plugin.json` + `.claude-plugin/marketplace.json` for
-  installation via `claude plugin install devrites@devrites-marketplace`.
-- `SECURITY.md` with private vulnerability-reporting channel, supported
-  versions, disclosure window, and documented security model.
-- `CODE_OF_CONDUCT.md` (Contributor Covenant 2.1).
-- `CODEOWNERS` mandating review on `pack/`, `scripts/`, `install.sh`,
-  `uninstall.sh`, `.claude-plugin/`.
-- Engineering rules ship under `pack/.claude/rules/` (16 rule files +
-  `README.md` index) and are autoloaded natively by Claude Code from
-  `.claude/rules/core.md`; on-demand files are read by the phase that needs
-  them. No carrier skill — `core.md` is always-on, the rest progressive-disclosure.
-- `pack/.claude/agents/devrites-simplifier-reviewer.md` (new) to complete
-  the audit-skill subagent set; existing
-  `devrites-security-auditor.md` + `devrites-performance-reviewer.md`
-  reused for security + performance audits.
-- `## Common Rationalizations` (4-6 rite-specific excuse/rebuttal rows) and
-  `## Red Flags` (3-5 bullets) sections on `rite-spec` and `rite-define`
-  SKILL.md (rollout continues — remaining `rite-*` skills carry the universal
-  anti-pattern table at `pack/.claude/rules/anti-patterns.md` plus
-  per-phase `reference/anti-patterns.md`).
-- Universal anti-rationalization table + red-flags list in
-  `pack/.claude/rules/anti-patterns.md`; minimal always-on subset in
-  `pack/.claude/rules/core.md`.
-- Trigger evals under `evals/<skill>.json` (20 should/should-not-trigger
-  queries per public `rite-*` skill) and `scripts/run-evals.sh`.
-- `scripts/devrites-detect.sh` — deterministic anti-slop regex detector
-  (25-30 rules across UI + code anti-slop).
-- `.github/workflows/ci.yml` running `validate.sh`, install/uninstall
-  smoke, fixture install, commitlint, eval suite on every PR.
-- UI numerical bar — OKLCH-only, 4pt scale, 100/300/500 motion durations
-  + 75 % exit, three-axis dark-mode compensation, fluid-vs-fixed type
-  scale, container queries, semantic z-index — extended into
-  `pack/.claude/skills/devrites-frontend-craft/reference/quality-standards.md`.
-- Extended anti-AI-slop list (side-stripe borders, em-dash overuse, pure
-  `#000` / `#fff`, all-CAPS body, extended reflex-font reject list) in
-  `pack/.claude/skills/rite-polish/reference/anti-ai-slop.md`.
-- `NEVER` lists on the UI discipline references (`shape.md`, `craft.md`,
-  `design-references.md`).
-- `docs/architecture.md`, `docs/command-map.md`, `docs/usage.md` now ship
-  with the repo (`docs/` un-gitignored; `docs/internal/` is the new home
-  for the previously-local research / development notes).
-- Interactive **type-GO** confirmation in `/rite-seal` before irreversible
-  git actions (commit, push, tag) — keeps auto-trigger UX while preventing
-  accidental side effects.
-
-### Changed
-
-- `/rite-review` output now uses the explicit
-  `Critical / Important / Suggestion / Nit / FYI` severity vocabulary.
-- `/rite-seal` gate logic: `Critical == 0` proceeds, `Critical > 0` blocks,
-  `Important > 0` triggers an interactive `y/N` confirmation. No numeric
-  composite score (avoids Goodhart on a self-scoring agent).
-- Audit skills consolidated into a single `devrites-audit` skill that
-  dispatches the security / perf / simplify reviewer subagent on an axis
-  argument — replaces the earlier per-axis skills. Migration from
-  `context: fork` to `Task`-tool subagent dispatch closes Anthropic bug
-  [#49559](https://github.com/anthropics/claude-code/issues/49559)
-  (`context: fork` silently inline under plugin install) and matches the
-  `devrites-doubt` → `devrites-doubt-reviewer.md` pattern already in use.
-- Public `rite-*` skill descriptions rewritten with explicit
-  user-language trigger phrases + `Not for…` negative-scope clauses.
-- README install section now documents the plugin path alongside the bash
-  installer. Added a "Security model" section pointing at `SECURITY.md`.
-- License clarified: personal use and plugin-marketplace **listing** are
-  permitted without approval; redistribution / mirroring / commercial
-  / organizational use still requires approval.
-
-## [0.1.0] — unreleased
-
-Initial public release line. Tag will be cut once the items above land on
-`main`.
-
-[Unreleased]: https://github.com/ViktorsBaikers/DevRites/compare/v0.1.0...HEAD
-[0.1.0]: https://github.com/ViktorsBaikers/DevRites/releases/tag/v0.1.0

package/NOTICE.md

@@ -4,7 +4,7 @@ DevRites is an original work. It vendors **no third-party code** — its skills,
 subagents, and engineering rules are written for DevRites.
 
 The Nielsen heuristics 0–4 scoring rubric in
-`rite-review/reference/nielsen-heuristics.md` references the heuristics themselves
+`pack/.claude/skills/rite-review/reference/nielsen-heuristics.md` references the heuristics themselves
 (Jakob Nielsen, 1994 / 2020 — public usability canon).
 
 ## License
@@ -12,7 +12,8 @@ Free to use. Modification is permitted **for personal use**; any other use —
 distributing it, distributing modified versions, or commercial/organizational use —
 requires **approval on request**. See [`LICENSE`](LICENSE).
 
-## Claude Code
-DevRites is a skills pack for **Claude Code** and follows its skills, subagents, and
-rules conventions. "Claude" and "Claude Code" are products of Anthropic; DevRites is
-independent and unaffiliated.
+## Host runtimes
+DevRites integrates with **Claude Code** and **Codex** and follows each host's
+project-local skill, agent, and instruction conventions. "Claude" and "Claude
+Code" are products of Anthropic; "Codex" is a product of OpenAI. DevRites is
+independent and unaffiliated with Anthropic and OpenAI.