create-sdd-project 0.9.7 → 0.9.9

package/README.md

@@ -294,19 +294,27 @@ SDD DevFlow combines three proven practices:
 | `qa-engineer` | Edge cases, spec verification | 5 |
 | `database-architect` | Schema design, optimization | Any |
 
-### 3 Skills (Slash Commands)
+### 4 Skills (Slash Commands)
 
 | Skill | Trigger | What it does |
 |-------|---------|-------------|
 | `development-workflow` | `start task F001`, `next task`, `add feature` | Orchestrates the complete 7-step workflow |
 | `bug-workflow` | `report bug`, `fix bug`, `hotfix needed` | Bug triage, investigation, and resolution |
 | `project-memory` | `set up project memory`, `log a bug fix` | Maintains institutional knowledge |
+| `health-check` | `health check`, `project health` | Quick scan: tests, build, specs sync, secrets, docs freshness |
+
+### 2 Custom Commands
+
+| Command | What it does |
+|---------|-------------|
+| `/review-plan` | Sends Implementation Plan to external AI models (Codex CLI, Gemini CLI) for independent critique |
+| `/context-prompt` | Generates a context recovery prompt after `/compact` with Workflow Recovery to prevent checkpoint skipping |
 
 ### Plan Quality
 
 Every Standard/Complex feature plan goes through a **built-in self-review** (Step 2.4) where the agent re-reads its own plan and checks for errors, vague steps, wrong assumptions, and over-engineering before requesting approval.
 
-For additional confidence, the optional `/review-plan` command sends the plan to an external AI model (Codex CLI, Gemini CLI, or Claude Code) for independent critique — catching blind spots that same-model review misses.
+For additional confidence, the optional `/review-plan` command sends the plan to external AI models (Codex CLI and/or Gemini CLI in parallel) for independent critique — catching blind spots that same-model review misses.
 
 ### Workflow (Steps 0–6)
 
@@ -336,6 +344,16 @@ For additional confidence, the optional `/review-plan` command sends the plan to
 
 Quality gates (tests, lint, build, validators) **always run** regardless of level.
 
+### Merge Checklist (B+D Mechanism)
+
+Every ticket includes a `## Merge Checklist Evidence` table that the agent must fill before requesting merge approval. This mechanism:
+
+- **Survives context compaction** — the ticket is always re-read via product tracker, so the empty evidence table acts as a persistent reminder
+- **Forces sequential execution** — agent must read `references/merge-checklist.md`, execute 9 actions (0–8), and record evidence
+- **Works at all tiers** — Simple tasks get a lite ticket with the same evidence table
+
+Validated across 16+ features with 87% first-attempt pass rate (failures led to iterative improvements in v0.8.7–v0.9.8).
+
 ### Project Memory
 
 Tracks institutional knowledge across sessions in `docs/project_notes/`:
@@ -399,14 +417,18 @@ project/
 │   │   ├── development-workflow/        # Main task workflow (Steps 0-6)
 │   │   │   └── references/              # Templates, guides, examples
 │   │   ├── bug-workflow/                # Bug triage and resolution
+│   │   ├── health-check/               # Project health diagnostics
 │   │   └── project-memory/              # Memory system setup
+│   ├── commands/                        # Custom slash commands
+│   │   ├── review-plan.md              # Cross-model plan review
+│   │   └── context-prompt.md           # Post-compact context recovery
 │   ├── hooks/quick-scan.sh              # Post-developer quality scan
 │   └── settings.json                    # Shared hooks (git-tracked)
 │
 ├── .gemini/
 │   ├── agents/                          # 9 agents (Gemini format)
-│   ├── skills/                          # Same 3 skills
-│   ├── commands/                        # Slash command shortcuts
+│   ├── skills/                          # Same 4 skills
+│   ├── commands/                        # Slash commands (workflow + review + context)
 │   └── settings.json                    # Gemini configuration
 │
 ├── ai-specs/specs/
@@ -462,11 +484,11 @@ cp -r node_modules/create-sdd-project/template/ /path/to/your-project/
 
 ## Roadmap
 
+- **PM Agent + L5 Autonomous**: AI-driven feature orchestration — sequential feature loop with automatic checkpoint approval and session state persistence
+- **`/review-project` command**: Comprehensive project review using multiple AI models in parallel (Claude + Gemini + Codex) with consolidated action plan
 - **Monorepo improvements**: Better support for pnpm workspaces and Turbo
-- **SDD Upgrade/Migration**: Version bumps for projects already using SDD
 - **Retrofit Testing**: Automated test generation for existing projects with low coverage
 - **Agent Teams**: Parallel execution of independent tasks
-- **PM Agent + L5 Autonomous**: AI-driven feature orchestration with human review at milestone boundaries
 
 ## Contributing
 

package/lib/config.js

@@ -106,6 +106,7 @@ const TEMPLATE_AGENTS = [
 // Template-provided command files (for upgrade: detect custom commands)
 const TEMPLATE_COMMANDS = [
   'review-plan.md',
+  'context-prompt.md',
 ];
 
 module.exports = {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "create-sdd-project",
-  "version": "0.9.7",
+  "version": "0.9.9",
   "description": "Create a new SDD DevFlow project with AI-assisted development workflow",
   "bin": {
     "create-sdd-project": "bin/cli.js"

package/template/.claude/commands/context-prompt.md

@@ -0,0 +1,44 @@
+Generate a complete context recovery prompt to paste after /compact or at the start of a new session.
+
+## What to include
+
+The prompt must contain EVERYTHING needed for a new session (or post-compact) to continue work without information loss:
+
+1. **Project state**: current branch, last commit, working tree status, SDD version
+2. **Workflow**: SDD development workflow, autonomy level, branching strategy
+3. **Active feature**: current step in the 6-step workflow, what has been done, what remains
+4. **Epics and progress**: table with status of each epic
+5. **Backlog**: pending features with priorities and dependencies
+6. **Infrastructure**: key modules, endpoints, schemas, with exact file paths
+7. **Tests**: total count, files, lint/build/tsc status
+8. **Key files to read**: list of files the new session must read first
+9. **Next action**: the specific next thing to do, with context
+10. **User notes**: any decisions or constraints communicated by the user
+
+### Workflow Recovery (CRITICAL)
+
+This section prevents the agent from losing track of the development process after /compact:
+
+11. **Current workflow step**: Which of the 6 steps (Spec, Setup, Plan, Implement, Finalize, Review) is active
+12. **Pending checkpoints**: Which approvals remain (Spec, Ticket, Plan, Commit, Merge)
+13. **Merge checklist reminder**: If at Step 5 or later, explicitly state: "Before requesting merge approval, you MUST read `references/merge-checklist.md` and execute ALL actions (0-8). Fill the `## Merge Checklist Evidence` table in the ticket with real evidence for each action."
+14. **Step order reminder**: "After commit+PR, run code-review-specialist and qa-engineer (Step 5), then execute merge-checklist actions. Do NOT request merge approval without completing the checklist."
+
+## How to generate
+
+- Read `docs/project_notes/product-tracker.md` (Active Session + Features tables + Completion Log)
+- Read `docs/project_notes/decisions.md` (recent ADRs)
+- Read `docs/project_notes/key_facts.md` (stack, components)
+- Read the current ticket in `docs/tickets/` if a feature is active
+- Run `git log --oneline -5` and `git status` for current state
+- Read `.sdd-version` for SDD DevFlow version
+
+## Output format
+
+Structured markdown with tables, ready to paste directly as the first message of a new session. The prompt should be self-contained — the receiving agent should not need to ask clarifying questions.
+
+## Context
+
+- Users work remotely with long sessions — interruptions are costly
+- After /review-plan + plan approval, context is typically at 50%+ usage, making /compact necessary before implementation
+- The workflow recovery section is essential: without it, the agent may skip merge checklist actions after /compact

package/template/.gemini/commands/context-prompt-instructions.md

@@ -0,0 +1,46 @@
+## Context Prompt — Instructions
+
+Generate a complete context recovery prompt to paste after compaction or at the start of a new session.
+
+### What to include
+
+The prompt must contain EVERYTHING needed for a new session (or post-compact) to continue work without information loss:
+
+1. **Project state**: current branch, last commit, working tree status, SDD version
+2. **Workflow**: SDD development workflow, autonomy level, branching strategy
+3. **Active feature**: current step in the 6-step workflow, what has been done, what remains
+4. **Epics and progress**: table with status of each epic
+5. **Backlog**: pending features with priorities and dependencies
+6. **Infrastructure**: key modules, endpoints, schemas, with exact file paths
+7. **Tests**: total count, files, lint/build/tsc status
+8. **Key files to read**: list of files the new session must read first
+9. **Next action**: the specific next thing to do, with context
+10. **User notes**: any decisions or constraints communicated by the user
+
+#### Workflow Recovery (CRITICAL)
+
+This section prevents the agent from losing track of the development process after compaction:
+
+11. **Current workflow step**: Which of the 6 steps (Spec, Setup, Plan, Implement, Finalize, Review) is active
+12. **Pending checkpoints**: Which approvals remain (Spec, Ticket, Plan, Commit, Merge)
+13. **Merge checklist reminder**: If at Step 5 or later, explicitly state: "Before requesting merge approval, you MUST read `references/merge-checklist.md` and execute ALL actions (0-8). Fill the `## Merge Checklist Evidence` table in the ticket with real evidence for each action."
+14. **Step order reminder**: "After commit+PR, run code-review-specialist and qa-engineer (Step 5), then execute merge-checklist actions. Do NOT request merge approval without completing the checklist."
+
+### How to generate
+
+- Read `docs/project_notes/product-tracker.md` (Active Session + Features tables + Completion Log)
+- Read `docs/project_notes/decisions.md` (recent ADRs)
+- Read `docs/project_notes/key_facts.md` (stack, components)
+- Read the current ticket in `docs/tickets/` if a feature is active
+- Run `git log --oneline -5` and `git status` for current state
+- Read `.sdd-version` for SDD DevFlow version
+
+### Output format
+
+Structured markdown with tables, ready to paste directly as the first message of a new session. The prompt should be self-contained — the receiving agent should not need to ask clarifying questions.
+
+### Context
+
+- Users work remotely with long sessions — interruptions are costly
+- After plan review + plan approval, context is typically at 50%+ usage, making compaction necessary before implementation
+- The workflow recovery section is essential: without it, the agent may skip merge checklist actions after compaction

package/template/.gemini/commands/context-prompt.toml

@@ -0,0 +1,2 @@
+description = "Generate a complete context recovery prompt to paste after compaction or at the start of a new session"
+prompt = "Read the file .gemini/commands/context-prompt-instructions.md and follow the instructions to generate a context recovery prompt."