openkash · one-kash · May 27, 2026 · May 27, 2026
diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
@@ -1,8 +1,8 @@
 {
   "name": "dev-workflow",
-  "version": "1.1.0",
-  "description": "Spec, Plan, Review, Implement, Verify — a complete AI development workflow with TDD, structured specifications, and independent review agents.",
-  "keywords": ["tdd", "specification", "code-review", "workflow", "quality"],
+  "version": "2.0.0",
+  "description": "Spec, Plan, Review, Implement, Verify — a complete AI development workflow with test-driven implementation, structured specifications, and independent review agents.",
+  "keywords": ["tdd", "implement", "specification", "code-review", "workflow", "quality"],
   "homepage": "https://github.com/openkash/ai-agent-dev-workflow",
   "author": {
     "name": "openkash"
@@ -14,9 +14,9 @@
       "description": "Structured specification with user stories, acceptance criteria, and interactive clarification"
     },
     {
-      "name": "tdd",
-      "path": "skills/tdd",
-      "description": "Test-driven development with chunk decomposition, JSON tracker, and 8-point quality checklist"
+      "name": "implement",
+      "path": "skills/implement",
+      "description": "Test-driven implementation with chunk decomposition, JSON tracker, and 8-point quality checklist"
     }
   ]
 }
diff --git a/.github/workflows/sync-skills.yml b/.github/workflows/sync-skills.yml
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -5,6 +5,29 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.0.0] - 2026-05-27
+
+### Changed (breaking)
+
+- Renamed `/tdd` to `/implement`. The slash command is now `/implement <feature>`. Speckit command ID is `speckit.dev-workflow.implement` (alias `speckit.implement`); the previous `speckit.dev-workflow.tdd` ID no longer resolves. Skill directory moved from `skills/tdd/` to `skills/implement/`.
+- Tracker filename convention is now `impl-tracker-<feature>.json`. The previous `tdd-tracker-*.json` prefix is no longer recognized by review-plan / review-impl.
+- Phase 6 parallel reviewer is now `/code-review` (read-only) instead of `/simplify` (which writes to the tree). Running a writer concurrently with review-impl was racy; the review pair is now both read-only.
+- This repo is now the only distribution channel. `openkash/ai-agent-spec-skill` and `openkash/ai-agent-tdd-skill` are deprecated; the spec-sync workflow and `sync.sh` script have been removed. Install via the Claude Code plugin or copy directly from this repo.
+
+### Added
+
+- UI state-holder vs rendering-only distinction in the test-strategy table. UI chunks that introduce `mutableStateOf`, `LaunchedEffect`/`useEffect`, hoisted state, or input transformation must extract a state holder and unit-test its transitions. Pure prop/callback threading with no logic keeps the "no test" exemption.
+- Both review-plan and review-impl now check the state-holder exemption explicitly. A chunk that claims "Pure UI — no test" while introducing a holder is flagged as a FAIL.
+- Small-feature shortcut keeps both review gates. Only Plan Mode (Phase 2.4) and chunk decomposition (Phase 2.1-2.2) collapse for 1-chunk features; review-plan and review-impl + /code-review still run.
+
+### Migration
+
+- Replace any saved `/tdd <feature>` invocations with `/implement <feature>`.
+- Rename existing in-flight tracker files from `tdd-tracker-*.json` to `impl-tracker-*.json` if you want them picked up by the agents on resumption.
+- Update any project-level docs or onboarding notes that reference `skills/tdd/` to `skills/implement/`.
+- If you customised `skills/tdd/PROJECT.md`, copy it to `skills/implement/PROJECT.md` before pulling — the directory rename means the old path will not exist after the update.
+- The `resume` field on tracker chunks has been removed; per-chunk pattern hints and pitfalls now live in the optional `notes` field.
+
 ## [1.1.0] - 2026-04-12
 
 ### Changed
@@ -32,5 +55,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Example configs for Android/Kotlin, TypeScript/Node, Python/Django, Python/pytest, Rust/CLI, Rust/Cargo
 - Session resumption via JSON tracker with per-chunk resume fields
 
+[2.0.0]: https://github.com/openkash/ai-agent-dev-workflow/releases/tag/v2.0.0
 [1.1.0]: https://github.com/openkash/ai-agent-dev-workflow/releases/tag/v1.1.0
 [1.0.0]: https://github.com/openkash/ai-agent-dev-workflow/releases/tag/v1.0.0
diff --git a/PRIVACY.md b/PRIVACY.md
@@ -9,7 +9,7 @@ This plugin collects no data. It consists entirely of markdown files (skills, ag
 
 ## What This Plugin Does
 
-- Provides two skills (`/spec`, `/tdd`) that guide Claude Code through structured development workflows
+- Provides two skills (`/spec`, `/implement`) that guide Claude Code through structured development workflows
 - Provides two review agents (`review-plan`, `review-impl`) that evaluate plans and implementations
 - All processing happens locally in your Claude Code session
 

diff --git a/README.md b/README.md
@@ -11,21 +11,21 @@ specify, plan, review, implement, verify.
 ```
 /spec            Clarify WHAT to build (user stories, acceptance criteria)
     |
-/tdd             Plan HOW to build it (chunks, dependencies, tracker)
+/implement       Plan HOW to build it (chunks, dependencies, tracker)
     |
     |--- GATE: tracker file triggers review-plan automatically
     v
 review-plan      Independent 8-point plan review (blocks implementation)
     |
-/tdd             Implement with TDD (failing test -> code -> pass)
+/implement       Implement test-first (failing test -> code -> pass)
     |
     |--- GATE: all chunks complete triggers parallel review
     v
 review-impl      Verifies implementation matches plan ---+
-/simplify        Reviews code reuse and quality ----------+--- run in parallel
+/code-review     Reviews code reuse and quality ---------+--- run in parallel
 ```
 
-`/tdd` handles both planning and implementation. Reviews are
+`/implement` handles both planning and implementation. Reviews are
 **artifact-triggered gates**, not optional suggestions. The tracker
 file triggers review-plan; chunk completion triggers review-impl.
 Both agents run in isolated context with no memory of writing the
@@ -34,22 +34,22 @@ code, so they evaluate honestly.
 ## Why This Workflow
 
 > [!TIP]
-> - **Reviews are gates, not suggestions.** The tracker file triggers review-plan before coding starts. Chunk completion triggers review-impl + /simplify in parallel. No manual invocation needed, no option to skip.
+> - **Reviews are gates, not suggestions.** The tracker file triggers review-plan before coding starts. Chunk completion triggers review-impl + /code-review in parallel. No manual invocation needed, no option to skip.
 > - **The reviewer didn't write the code.** Review agents spawn in a fresh context, separate from the author. Bias eliminated by architecture, not instruction.
 > - **Scope is checked in both directions.** review-plan verifies every criterion has a chunk (completeness) AND every chunk serves a criterion (containment). Catches scope expansion before it becomes code.
 > - **Design bugs are caught before coding starts.** 8-point plan review before any code is written. Fixing a wrong abstraction in a plan costs minutes. In code, hours.
-> - **Specs are created, not just read.** `/spec` writes user stories with Given/When/Then acceptance criteria, then `/tdd` implements them. From "what should we build?" to passing tests.
-> - **Work survives context resets.** JSON tracker with per-chunk `resume` fields and `plan_review` verdict tells a new session exactly where to pick up.
+> - **Specs are created, not just read.** `/spec` writes user stories with Given/When/Then acceptance criteria, then `/implement` implements them. From "what should we build?" to passing tests.
+> - **Work survives context resets.** JSON tracker with per-chunk `name`, `files_*`, `tdd`, `acceptance_criteria` (plus optional `notes`) and the top-level `plan_review` verdict tells a new session exactly where to pick up.
 > - **One install, complete workflow.** Four pieces that work as a pipeline. No separate tools to discover, install, and wire.
 
 ## What's Included
 
 | Piece | Type | Invocation | Purpose |
 |-------|------|-----------|---------|
 | [spec](skills/spec/SKILL.md) | Skill | `/spec <feature>` | User stories, acceptance criteria, edge cases |
-| [tdd](skills/tdd/SKILL.md) | Skill | `/tdd <feature>` | Chunk decomposition, TDD cycle, quality checklist |
-| [review-plan](agents/review-plan.md) | Agent | Triggered by /tdd | 8-point plan review with scope containment check |
-| [review-impl](agents/review-impl.md) | Agent | Triggered by /tdd | Checks code matches the plan and passes acceptance criteria |
+| [implement](skills/implement/SKILL.md) | Skill | `/implement <feature>` | Chunk decomposition, test-driven cycle, quality checklist |
+| [review-plan](agents/review-plan.md) | Agent | Triggered by /implement | 8-point plan review with scope containment check |
+| [review-impl](agents/review-impl.md) | Agent | Triggered by /implement | Checks code matches the plan and passes acceptance criteria |
 
 ## Installation
 
@@ -69,35 +69,27 @@ and both agents are available to use.
 git clone https://github.com/openkash/ai-agent-dev-workflow.git /tmp/dev-workflow
 
 # Skills
-mkdir -p .claude/skills/spec .claude/skills/tdd
+mkdir -p .claude/skills/spec .claude/skills/implement
 cp /tmp/dev-workflow/skills/spec/SKILL.md .claude/skills/spec/
 cp -r /tmp/dev-workflow/skills/spec/references/ .claude/skills/spec/
 cp -r /tmp/dev-workflow/skills/spec/project-configs/ .claude/skills/spec/
-cp /tmp/dev-workflow/skills/tdd/SKILL.md .claude/skills/tdd/
-cp -r /tmp/dev-workflow/skills/tdd/references/ .claude/skills/tdd/
-cp -r /tmp/dev-workflow/skills/tdd/project-configs/ .claude/skills/tdd/
+cp /tmp/dev-workflow/skills/implement/SKILL.md .claude/skills/implement/
+cp -r /tmp/dev-workflow/skills/implement/references/ .claude/skills/implement/
+cp -r /tmp/dev-workflow/skills/implement/project-configs/ .claude/skills/implement/
 
 # Agents
 mkdir -p .claude/agents
 cp /tmp/dev-workflow/agents/*.md .claude/agents/
 ```
 
-### Individual skills (standalone repos)
-
-If you only want the spec skill without the full workflow:
-
-- **Spec only:** [openkash/ai-agent-spec-skill](https://github.com/openkash/ai-agent-spec-skill)
-
-> **Note:** [ai-agent-tdd-skill](https://github.com/openkash/ai-agent-tdd-skill) is deprecated. This repo is the canonical source for the TDD skill, which depends on the review agents for its gate system.
-
 ## Configuration
 
 Each skill reads a `PROJECT.md` in its directory for project-specific
 settings (build commands, test commands, standards).
 
 ```bash
 # Copy templates and customize
-cp /tmp/dev-workflow/skills/tdd/PROJECT.md .claude/skills/tdd/PROJECT.md
+cp /tmp/dev-workflow/skills/implement/PROJECT.md .claude/skills/implement/PROJECT.md
 cp /tmp/dev-workflow/skills/spec/PROJECT.md .claude/skills/spec/PROJECT.md
 ```
 
@@ -113,25 +105,25 @@ Example configs for common stacks are in `skills/*/project-configs/`.
 Produces `docs/specs/rate-limiting.md` with user stories and acceptance criteria.
 
 ```
-/tdd implement rate limiting (spec: docs/specs/rate-limiting.md)
+/implement rate limiting (spec: docs/specs/rate-limiting.md)
 ```
-The TDD skill decomposes the work into chunks, creates a tracker, and
-asks for approval. The tracker file automatically triggers the
-review-plan gate:
+The /implement skill decomposes the work into chunks, creates a
+tracker, and asks for approval. The tracker file automatically
+triggers the review-plan gate:
 
 ```
 GATE: tracker triggers review-plan agent (blocks Phase 3)
 -> Returns structured verdict (PASS/WARN/FAIL per criterion)
 -> FAIL: update plan and re-run. PASS: proceed to implementation.
 ```
 
-After all chunks are implemented, the TDD skill triggers the
+After all chunks are implemented, /implement triggers the
 implementation review gate (both run in parallel):
 
 ```
 GATE: all chunks complete triggers parallel review
 -> review-impl agent verifies plan conformance and acceptance criteria
--> /simplify reviews code reuse and quality
+-> /code-review reviews code reuse and quality
 -> Address any FAIL findings before completing.
 ```
 
@@ -141,7 +133,7 @@ The agents work independently of the skills:
 
 ```
 Use the review-plan agent to review my-plan.md
-Use the review-impl agent to check the implementation against docs/tdd-tracker.json
+Use the review-impl agent to check the implementation against docs/impl-tracker.json
 ```
 
 ## File Structure
@@ -156,15 +148,14 @@ ai-agent-dev-workflow/
 │   │   ├── PROJECT.md           # Template - copy and customize
 │   │   ├── references/          # Spec template, taxonomy, validation
 │   │   └── project-configs/     # Example configs per stack
-│   └── tdd/                     # TDD skill
-│       ├── SKILL.md             # Core TDD process
+│   └── implement/               # Implementation skill
+│       ├── SKILL.md             # Core implementation process
 │       ├── PROJECT.md           # Template - copy and customize
 │       ├── references/          # Chunk template, tracker schema, checklist
 │       └── project-configs/     # Example configs per stack
 ├── agents/
 │   ├── review-plan.md           # Plan review agent (8-point)
 │   └── review-impl.md          # Implementation review agent (8-point)
-├── sync.sh                      # Sync to individual repos (maintainer use)
 ├── LICENSE                      # Apache 2.0
 └── README.md
 ```
@@ -179,7 +170,7 @@ The review agents are subagents (not slash-command skills) by design:
 |--------|----------------------|-------------------|
 | **Evaluator independence** | Runs in same context as author, biased toward own work | Fresh context, no memory of writing the plan/code |
 | **Context pollution** | Review details consume main window | Isolated, returns summary only |
-| **Automation** | User must remember to invoke | /tdd gates trigger automatically from artifacts |
+| **Automation** | User must remember to invoke | /implement gates trigger automatically from artifacts |
 | **Parallel execution** | Sequential in main context | Can run both reviews in parallel |
 | **Standalone use** | User invokes via `/review-plan` | User invokes via natural language ("review this plan") |