opencode-sdlc-plugin 0.1.0

package/package.json

@@ -0,0 +1,56 @@
+{
+  "name": "opencode-sdlc-plugin",
+  "version": "0.1.0",
+  "description": "OpenCode SDLC workflow plugin with event modeling, ADRs, and TDD orchestration",
+  "type": "module",
+  "bin": {
+    "opencode-sdlc-plugin": "dist/cli/index.js"
+  },
+  "exports": {
+    ".": "./dist/index.js",
+    "./plugin": "./dist/plugin/index.js"
+  },
+  "publishConfig": {
+    "access": "public"
+  },
+  "files": [
+    "dist",
+    "commands",
+    "skills",
+    "agents",
+    "config"
+  ],
+  "scripts": {
+    "build": "tsup",
+    "dev": "tsup --watch",
+    "lint": "eslint .",
+    "typecheck": "tsc --noEmit",
+    "test": "vitest run",
+    "test:watch": "vitest",
+    "test:coverage": "vitest run --coverage",
+    "test:integration": "vitest run --config vitest.integration.config.ts",
+    "test:coverage:integration": "vitest run --coverage --config vitest.integration.config.ts"
+  },
+  "dependencies": {
+    "@inquirer/prompts": "^7.1.0",
+    "@opencode-ai/plugin": "^1.1.24",
+    "commander": "^12.1.0",
+    "fs-extra": "^11.2.0",
+    "kleur": "^4.1.5",
+    "ora": "^8.0.1",
+    "picomatch": "^4.0.2",
+    "zod": "^3.23.8"
+  },
+  "devDependencies": {
+    "@eslint/js": "^9.39.2",
+    "@types/fs-extra": "^11.0.4",
+    "@types/node": "^20.11.30",
+    "@types/picomatch": "^4.0.2",
+    "@vitest/coverage-v8": "^4.0.17",
+    "eslint": "^9.9.0",
+    "tsup": "^8.2.4",
+    "typescript": "^5.6.3",
+    "typescript-eslint": "^8.53.0",
+    "vitest": "^4.0.17"
+  }
+}

package/skills/adr-policy.md

@@ -0,0 +1,21 @@
+---
+description: ADR policy and ARCHITECTURE synthesis
+---
+
+# ADR Policy
+
+## ADR Isolation (CRITICAL)
+
+ADRs are archival only. Never reference ADRs in issues, PRs, reviews, or code. Use `docs/ARCHITECTURE.md`.
+
+## ADR Format
+
+- Context
+- Decision
+- Consequences
+
+## ARCHITECTURE.md
+
+- Synthesizes all accepted ADRs
+- Treated as the current state projection
+- Regenerated when ADRs change

package/skills/atomic-design.md

@@ -0,0 +1,39 @@
+---
+description: Atomic Design UI methodology
+---
+
+# Atomic Design
+
+Use for UI systems; skip for CLI tools, libraries, or API-only services.
+
+## Prerequisites
+
+- `docs/ARCHITECTURE.md` exists
+- Event model workflows include wireframes (if applicable)
+
+## Hierarchy
+
+| Level | Description | Examples |
+|-------|-------------|----------|
+| Atoms | Basic elements | Buttons, inputs, labels, icons |
+| Molecules | Small combinations | Form fields, cards |
+| Organisms | Complex components | Navigation, forms |
+| Templates | Page layouts | Dashboard, list view |
+| Pages | Assembled screens | Full flows |
+
+## Process
+
+1. Analyze wireframes for common patterns.
+2. Capture visual style preferences.
+3. Create design tokens (color, typography, spacing).
+4. Build hierarchy from atoms → molecules → organisms → templates.
+5. Map read models to component data requirements.
+
+## Artifacts Location
+
+- `docs/design-system/`
+
+## Integration
+
+- Uses event model wireframes.
+- Reviewed by `@ux`.

package/skills/debugging-protocol.md

@@ -0,0 +1,47 @@
+---
+description: Debugging protocol for investigation
+---
+
+# Debugging Protocol (MANDATORY)
+
+No fixes without root-cause investigation.
+
+## Before Debugging
+
+Always search memento:
+```
+/sdlc-recall "<error message> fix"
+```
+
+## Phase 1: Root Cause Investigation
+
+- Read full error output.
+- Reproduce consistently.
+- Trace failing data flow.
+- Check recent changes.
+
+## Phase 2: Pattern Analysis
+
+- Find working examples.
+- Compare line-by-line differences.
+- Identify dependencies.
+
+## Phase 3: Hypothesis and Testing
+
+- Form ONE hypothesis.
+- Test with ONE minimal change.
+- Confirm or refute before proceeding.
+
+## Phase 4: Implementation
+
+- Add/confirm failing test.
+- Fix root cause, not symptoms.
+- Verify with tests.
+
+## Escalation Rule
+
+After 3 failed fixes: stop and re-examine architecture.
+
+## Defense in Depth
+
+Add validation across layers (entry, business, environment, diagnostics).

package/skills/event-modeling.md

@@ -0,0 +1,40 @@
+---
+description: Event modeling methodology
+---
+
+# Event Modeling
+
+Follow Martin Dilger’s event sourcing methodology.
+
+## Core Principles
+
+- Events are immutable facts in past tense using business language.
+- Do not lose information; store what happened, not just current state.
+- Every read model attribute must trace back to an event.
+
+## Four Patterns
+
+1. **State Change**: Command → Event (state mutation)
+2. **State View**: Events → Read Model (query projection)
+3. **Automation**: Event → Process → Command → Event (background work)
+4. **Translation**: External data → Internal event (anti-corruption layer)
+
+## Mapping to Work Tracking (Non-Negotiable)
+
+| Event Modeling | GitHub Tracking |
+|---------------|-----------------|
+| Workflow | Epic (parent issue) |
+| Vertical Slice | Story issue (1:1) |
+| GWT scenarios | Acceptance criteria |
+
+## Artifacts Location
+
+- `docs/event_model/domain/overview.md`
+- `docs/event_model/workflows/<workflow>/overview.md`
+- `docs/event_model/workflows/<workflow>/slices/*.md`
+
+## Story Review Perspectives
+
+- `@story` for business value and slice thinness
+- `@architect` for feasibility and risks
+- `@ux` for user journey coherence

package/skills/git-spice.md

@@ -0,0 +1,44 @@
+---
+description: git-spice stacked PR workflow
+---
+
+# git-spice Workflow
+
+Use only if `features.gitSpice` is enabled.
+
+## When to Use
+
+- Dependent, stacked changes
+- Already on a gs-managed branch
+
+## When to Skip
+
+- Fresh work from main
+- Standalone changes
+
+## Detection
+
+```
+command -v gs && gs branch checkout 2>/dev/null
+```
+
+## Common Commands
+
+```
+gs branch create <branch-name>
+gs stack submit
+gs repo sync
+gs up
+gs down
+gs stack
+```
+
+## Decision Tree
+
+1. Is git-spice installed? If not, use regular git.
+2. Is branch gs-managed? If not, start with `gs branch create`.
+3. Ready to submit? Use `gs stack submit`.
+
+## Error Handling
+
+If gs fails: run `gs repo sync`, then fall back to regular git.

package/skills/github-issues.md

@@ -0,0 +1,44 @@
+---
+description: GitHub CLI extensions usage
+---
+
+# GitHub CLI Extensions (MANDATORY)
+
+Always prefer extensions over `gh api`.
+
+## Priority Order
+
+1. Extension commands
+2. Native `gh` commands
+3. `gh api` (last resort)
+
+## Extensions
+
+### gh-issue-ext
+- `gh issue-ext sub list <issue>`
+- `gh issue-ext sub add <parent> <child>`
+- `gh issue-ext blocking add <blocked> <blocker>`
+- `gh issue-ext blocking list <issue>`
+- `gh issue-ext branch create <issue>`
+
+### gh-project-ext
+- `gh project-ext ready`
+- `gh project-ext move <issue> "In Progress"`
+- `gh project-ext claim <issue>`
+
+### gh-pr-review
+- `gh pr-review review view --pr <id> --unresolved`
+- `gh pr-review comments reply --thread-id <id> --body "..."`
+- `gh pr-review threads resolve --thread-id <id>`
+
+## `gh api` Is a Last Resort
+
+Before using `gh api`:
+1. Check native `gh` commands
+2. Check installed extensions: `gh extension list`
+3. Search for extensions: `gh extension search <keywords>`
+4. Ask user whether to install or proceed
+
+## Task Management
+
+GitHub Issues/Projects are the source of truth. Local todos are session-only.

package/skills/memory-protocol.md

@@ -0,0 +1,41 @@
+---
+description: Memento MCP memory protocol
+---
+
+# Memory Protocol (MANDATORY)
+
+Knowledge capture and retrieval are prime directives.
+
+## Memory Skills
+
+| Skill | Purpose | When to Use |
+|-------|---------|-------------|
+| `/sdlc-recall <query>` | Retrieve knowledge | Before ANY task, when errors occur |
+| `/sdlc-remember <what>` | Store discoveries | After solving problems or learning conventions |
+
+## Before Any Task
+
+Always run `/sdlc-recall` first. If you skip this, you are out of compliance.
+
+## When Commands Fail
+
+Search first, debug second:
+1. `/sdlc-recall "<error message> fix"`
+2. Apply known solution
+3. Store new solution with `/sdlc-remember`
+
+## During and After Work
+
+Store non-obvious insights:
+- Solutions found through trial and error
+- Project conventions or patterns
+- User preferences and workflow choices
+- Root-cause debugging insights
+
+## Subagent Responsibilities
+
+Subagents should use `mcp__memento__*` tools directly and record discoveries.
+
+## Before Session End
+
+Store unsaved discoveries prior to context compaction.

package/skills/orchestration.md

@@ -0,0 +1,118 @@
+---
+description: SDLC orchestration and delegation rules
+---
+
+# SDLC Orchestration Rules
+
+The main conversation is an orchestrator only. It coordinates work but never edits files directly.
+
+## Core Protocol: Skill Enforcement
+
+Before ANY task, invoke skill enforcement (1% rule, mandatory invocations, rationalization red flags).
+
+## ADR Isolation Principle (CRITICAL)
+
+ADRs are archival documents. NEVER reference ADRs in issues, PRs, code comments, reviews, or slice docs. Always reference `docs/ARCHITECTURE.md` instead. Only consult ADRs when reconsidering architecture decisions.
+
+## File Operations (Mandatory Delegation)
+
+The orchestrator MUST NEVER use Write/Edit. Delegate all file changes to agents.
+
+### Agent Selection Hierarchy
+
+| File Type | Agent | Notes |
+|-----------|-------|-------|
+| Test files | `@red` | All test code, fixtures, helpers |
+| Production code | `@green` | Minimal implementation only |
+| Domain types | `@domain` | Value objects, types, invariants |
+| ADRs | `@adr` | `docs/adr/*.md` |
+| Architecture | `@architect` | `docs/ARCHITECTURE.md` |
+| Event model docs | `@exploration`, `@discovery`, `@workflowDesigner`, `@gwt`, `@modelChecker` | `docs/event_model/**/*` |
+| Everything else | `@fileUpdater` | Config, docs, scripts |
+
+### Update Request Detection
+
+| Request Pattern | Agent |
+|-----------------|-------|
+| "Add a test" | `@red` |
+| "Make the test pass" | `@green` |
+| "Add a domain type" | `@domain` |
+| "Write ADR" | `@adr` |
+| "Update architecture" | `@architect` |
+| "Update docs/config" | `@fileUpdater` |
+
+## TDD Workflow (Mandatory)
+
+```
+RED → DOMAIN → GREEN → DOMAIN → REFACTOR
+```
+
+- One failing test per RED cycle.
+- DOMAIN review is mandatory after RED and after GREEN.
+- No exceptions for “trivial” changes.
+
+### Anti-Patterns (Violations)
+
+- “Quick fix” → violates TDD.
+- “Just a UI tweak” → still requires domain review.
+- “One-line change” → still requires full cycle.
+
+### Pre-Edit Checklist
+
+- Is this a test file? → `@red`
+- Is this production code? → `@green`
+- Is this a domain type? → `@domain`
+- Has domain review happened before GREEN?
+
+## Fresh Context Protocol (CRITICAL)
+
+Agents have ZERO context. Always provide:
+- File paths and excerpts
+- Current test name
+- Failure output
+- Acceptance criteria
+- Relevant domain types
+
+## Invocation Gate Protocol
+
+### Red Agent
+```
+RED_CONTEXT: FIRST_TEST | CONTINUING
+ACCEPTANCE_CRITERIA:
+- <criterion>
+```
+
+### Domain Agent
+```
+DOMAIN_CONTEXT: AFTER_RED | AFTER_GREEN
+PHASE_COMPLETE:
+- Test: <name>
+- Failure/Result: <details>
+```
+
+### Green Agent
+```
+RED_PHASE_COMPLETE:
+- Test: <name>
+- Failure: <error>
+DOMAIN_CHECK_PASSED:
+- Types: <list> OR "None"
+```
+
+## Agent Debate Protocol
+
+- Domain agent has veto power.
+- If dispute persists for 2 rounds, escalate to user.
+
+## Parallel Development (Worktrees)
+
+If `config.git.worktrees` is true:
+- Use `git worktree list` to choose context.
+- Only run parallel slices with defined interfaces.
+
+## Code Review Gate
+
+Three-stage review before PR:
+1. Spec compliance
+2. Code quality
+3. Domain integrity (compile-time enforcement audit)

package/skills/skill-enforcement.md

@@ -0,0 +1,56 @@
+---
+description: Skill invocation enforcement rules
+---
+
+# Skill Invocation Enforcement (MANDATORY)
+
+## The 1% Rule
+
+If there is even a 1% chance a skill applies, you MUST invoke it.
+
+## Mandatory Invocations
+
+| Before... | ALWAYS invoke... | Even if... |
+|-----------|------------------|-----------|
+| ANY task | `/sdlc-recall` | "It's a simple change" |
+| Writing ANY test | `@red` | "Just one test" |
+| Writing ANY production code | `@green` | "One line" |
+| Creating ANY type | `@domain` | "It's just a struct" |
+| After RED | `@domain` | "Not a domain concern" |
+| After GREEN | `@domain` | "Just UI" |
+| Debugging | `debugging-protocol` | "I know the fix" |
+| Finishing work | `/sdlc-remember` | "Not interesting" |
+
+## Rationalization Red Flags
+
+- "This is simple"
+- "Overkill"
+- "Just this once"
+- "No need for domain review"
+- "I'll check memory after"
+
+## Skill Priority Order
+
+1. Memory (`/sdlc-recall`)
+2. Process (debugging protocol, orchestration)
+3. Implementation (red/green/domain)
+4. Documentation (ADR, GWT)
+
+## Non-Negotiables
+
+- Memory recall before tasks
+- TDD delegation via agents
+- Domain review after RED and GREEN
+- Verification before completion
+- Memory storage after discoveries
+- ADR isolation
+
+## ADR Isolation (CRITICAL)
+
+Never reference ADRs in issues, PRs, reviews, or code. Use `docs/ARCHITECTURE.md`.
+
+## Self-Check Questions
+
+- Did I search memento first?
+- Am I delegating to the right agent?
+- Am I skipping a step due to time pressure?

package/skills/tdd-constraints.md

@@ -0,0 +1,63 @@
+---
+description: File ownership and TDD constraints
+---
+
+# Agent File Ownership Constraints
+
+Each SDLC agent has strict file boundaries. These are ABSOLUTE.
+
+## Ownership Matrix
+
+| File Pattern | Authorized Agent(s) | Others Blocked |
+|--------------|---------------------|----------------|
+| Test files (`*_test.*`, `*.test.*`, `tests/`, `__tests__/`, `spec/`) | `@red` | ✅ |
+| Production code (`src/`, `lib/`, `app/`) | `@green` | ✅ |
+| Domain types/models | `@domain` | ✅ |
+| ADRs (`docs/adr/*.md`) | `@adr` | ✅ |
+| `docs/ARCHITECTURE.md` | `@architect`, `@designFacilitator` | ✅ |
+| Event model docs (`docs/event_model/**/*`) | `@exploration`, `@discovery`, `@workflowDesigner`, `@gwt`, `@modelChecker` | ✅ |
+| Config/docs/scripts | `@fileUpdater` | - |
+
+## TDD Agents
+
+### `@red`
+CAN edit: test files, fixtures, helpers.
+CANNOT edit: production code, domain types, architecture.
+
+### `@green`
+CAN edit: production implementation only.
+CANNOT edit: test files, type definitions, event model docs.
+
+### `@domain`
+CAN edit: domain types, value objects, invariants.
+CANNOT edit: tests, implementation bodies, ADRs.
+
+## Architecture Agents
+
+### `@adr`
+CAN edit: `docs/adr/*.md` only.
+CANNOT edit: `docs/ARCHITECTURE.md` directly.
+
+### `@architect`
+CAN edit: `docs/ARCHITECTURE.md` only.
+CANNOT edit: ADRs, event model docs, code.
+
+### `@designFacilitator`
+CAN edit: `docs/ARCHITECTURE.md` only.
+CANNOT edit: ADRs, event model docs, code.
+
+## Event Modeling Agents
+
+### `@exploration`, `@discovery`, `@workflowDesigner`, `@gwt`, `@modelChecker`
+CAN edit: `docs/event_model/**/*`.
+CANNOT edit: ADRs, architecture, code.
+
+## Utility Agent
+
+### `@fileUpdater`
+CAN edit: config, scripts, general docs not owned by other agents.
+CANNOT edit: tests, production code, domain types, ADRs, architecture, event model docs.
+
+## If Blocked
+
+Stop immediately, return to the orchestrator, and request the proper agent.