create-byan-agent 2.42.0 → 2.45.0

package/CHANGELOG.md

@@ -9,6 +9,124 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [2.45.0] - 2026-07-15
+
+### Added - Plain-language guard for every agent (Mantra IA-26 "Parler Reel")
+
+- New team-wide rule: agents speak clear, coherent French to the user — no gratuitous
+  English when a French word exists ("redemarrer le conteneur", not "cutoff"), no raw
+  internal jargon (leaf/tier/downgrade/gate/inline/advisory), no misapplied metaphor
+  (you do not "forge" a token). A technical term with no French equivalent (commit,
+  cache, token) is kept but explained once. Test: the reader understands with no
+  dictionary. This is the sibling of IA-23 (no emoji) and applies to ALL agents.
+- Mechanism, in three layers, none of which re-generates an already-shown reply:
+  1. The rule reaches every agent: mantra `IA-26` (`mantras.yaml` + `mantras-sources.md`),
+     `.claude/rules/plain-language.md`, and a pointer in `CLAUDE.md`.
+  2. BYAN's own voice stays fresh: a line in the per-turn reminder
+     (`inject-voice-anchor.js`) + entries in `tao.md` Section 4 (Vocabulaire Interdit).
+  3. A forward net (non-blocking): the Stop hook `plain-language-check.js` spots the
+     known repeat-offenders in the finished reply and writes a one-turn flag under
+     `_byan-output/`; the next-turn reminder reads it, signals it in plain French, and
+     clears it. No blocking, no re-answer — the correction is carried to the next turn
+     (a blocking guard would force a costly regen and the user has already read the slip).
+- Core logic isolated in `.claude/hooks/lib/plain-language.js` (French-aware word
+  boundary so "metier"/"chantier" stay clear of the "tier" match; code spans stripped
+  before scanning). Tests: `.claude/__tests__/plain-language.test.js`. Full suite green
+  (2629 jest). Shipped via `install/templates/`.
+
+## [2.44.0] - 2026-07-15
+
+### Changed - Revived the Sonnet middle tier for native-workflow model routing
+
+- Native-workflow ANALYSIS leaves (score/rank/assess/design/nfr/coverage/
+  recommend/synthesize) now auto-route to the balanced tier (`model: 'sonnet'`)
+  instead of `deep` (inherit the session model). Before this, the Sonnet rung was
+  reachable only through the explicit `mech-` opt-in, so on an Opus (or high-effort)
+  session every analysis/verification/implementation leaf ran on Opus — the Sonnet
+  tier was effectively dead (ledger: recent workflows logged all-`inherit`).
+- This deliberately overrides the session model for analysis leaves; a genuinely
+  frontier analysis opts back to the session model with the `deep-` label prefix
+  (`deep-assess-architecture`), the mirror of the `mech-` opt-in. VERIFICATION and
+  IMPLEMENTATION are untouched — a wrong check or a code-write still inherits the
+  session model (no silent downgrade of the dangerous classes).
+- Propagated through the enforcement stack: `modelRoutingViolations` accepts sonnet
+  on an analysis leaf (its tier) and still blocks haiku (`analysis-below-tier`) and
+  any downgrade on a protected leaf; added a non-blocking `untiered-analysis`
+  advisory (surfaces an analysis leaf running deep, suggests `model: 'sonnet'`).
+  Files: `native-tiers.js` (tierFor + `DEEP_PREFIX` + `synthes` keyword),
+  `workflows-lint.js`, `bin/byan-lint-workflows.js`, native-workflows.md.
+- The 21 committed workflows still pass the contract; MCP 792 + jest 2606 green.
+
+## [2.43.0] - 2026-07-10
+
+### Removed - Purged the Gen1 `_bmad/` legacy layer
+
+- Removed the Gen1 `_bmad/` fallback from the layout resolver
+  (`src/byan-v2/lib/layout-resolver.js`): `resolveAgent` / `agentDirs` now resolve
+  Gen3 (`_byan/agent/<name>/`) and Gen2 (`_byan/agents/`, `_byan/<module>/agents/`)
+  only. A pre-Gen3 project on the raw `_bmad/` layout is migrated with
+  `scripts/migrate-bmad-to-byan.js` (agents land under `_byan/`, resolved by the
+  Gen2 branches). Added a regression test locking Gen1 resolution to null.
+- Deleted the frozen Gen1 content: `_bmad/` (677 files) and the dead shipped
+  template `install/templates/_bmad/` (17 files) that the installer does not copy
+  (it scaffolds `templates/_byan/`). Neither carried a live code path.
+- Removed the dead `marc` agent residues (a GitHub Copilot CLI specialist for the
+  already-removed Copilot platform, present only under `_bmad/`): the
+  `subagent-generator` model-map key, an `agent-launcher` jsdoc example, and the
+  `marc.md` entry in the migrator's agent list.
+- Migrated `_bmad` doc references to `_byan` (CLAUDE.md, QUICK-START,
+  GUIDE-UTILISATION, soul-activation, workers) and cleaned stale Gen1
+  comments / dead branches (agent-packager, bridge, cli-detector, index example).
+- Kept the migration back-compat intact: the webui legacy-install detection, the
+  `migrate-bmad-to-byan` script, `stub-sync`, and the `_bmad-output` user-data scan.
+
+### Note - Known follow-up
+
+- `install/install.sh` (a dormant legacy shell installer) still scaffolds a Gen1
+  `_bmad/` tree and promotes GitHub Copilot CLI. It ships but is not run by the
+  npx flow. It needs a dedicated rework to align with the Gen3-only, Copilot-free
+  product; deliberately left out of this purge to avoid touching an untested
+  installer path.
+
+## [2.42.1] - 2026-07-06
+
+### Changed - Excised the dead stub half of the yanstaller module
+
+- Removed six pure-stub files under `install/lib/yanstaller/` that had no caller
+  (`installer.js`, `validator.js`, `recommender.js`, `interviewer.js`,
+  `troubleshooter.js`, `wizard.js`). `installer.js` faked a successful install
+  without touching the disk; `validator.js` returned `passed: true` for its ten
+  checks without inspecting anything. The real installer is
+  `install/bin/create-byan-agent-v2.js`.
+- Dropped the half-implemented `install()`/`uninstall()` orchestrator from
+  `install/lib/yanstaller/index.js`. The live surface consumed by the CLI
+  (`update` / `rollback` / `backups` / `check`) and the web UI is untouched:
+  `update`, `rollback`, `listBackups`, plus the `detector` / `platformSelector`
+  / `updater` / `backuper` modules.
+- Removed the dead `yanstaller.install()` call from the web UI install path
+  (`install/src/webui/api.js`); it already had its own directory/config
+  fallback and had logged that call as a stub.
+- Added `install/__tests__/yanstaller/no-stub-guard.test.js`, a regression guard
+  that keeps the excised stubs from creeping back and locks the surviving public
+  surface. Net: -853 / +11 lines shipped to npm consumers.
+
+### Changed - Purged emoji from the installer CLI output (Mantra IA-23)
+
+- Removed every emoji from the 16 live installer files under `install/lib/` and
+  `install/bin/` (201 occurrences). Status glyphs became bracket tags matching
+  the existing `[DEBUG]` idiom: `[OK]` / `[INFO]` / `[WARN]` / `[ERROR]` / `[x]`;
+  decorative glyphs were dropped and the arrow `->` de-symbolized. Message text,
+  colors, and control flow are unchanged (a `git diff -w` is emoji-only).
+- Updated the two coupled tests in lockstep: `install/__tests__/utils/logger.test.js`
+  and `install/__tests__/integration/platform-integration.test.js`.
+
+### Removed - Dead Copilot-era task-tool-interface stub
+
+- Deleted `task-tool-interface.js` + `-mock.js` from both `src/byan-v2/dispatcher/`
+  and `install/src/byan-v2/dispatcher/`, plus their test. The module had no caller
+  outside its own test. Scrubbed the five stale `TaskToolInterface` doc references
+  it left behind (`task-router` JSDoc + `COMPLETION-REPORT.md`).
+
 ## [2.42.0] - 2026-07-06
 
 ### Fixed - Shipped Claude Code hooks no longer spam MODULE_NOT_FOUND

package/install/bin/byan-kanban.js

@@ -86,7 +86,7 @@ function renderStandups(entries) {
       const blockers = e.blockers && e.blockers.length > 0
         ? FG(COLORS.blocked, ` blockers=${e.blockers.join('|')}`)
         : '';
-      return `${e.timestamp} ${e.agent}: ${e.did || '-'}${blockers} → ${e.next || '-'}`;
+      return `${e.timestamp} ${e.agent}: ${e.did || '-'}${blockers} -> ${e.next || '-'}`;
     })
     .join('\n');
 }