reviewflow 3.0.0

package/templates/SKILL.md.template

@@ -0,0 +1,207 @@
+---
+name: review-{{PROJECT_TYPE}}
+description: Code review d'une MR/PR avec audits parallèles. Génère un rapport et le poste sur GitLab/GitHub.
+---
+
+# Code Review {{PROJECT_TYPE}}
+
+## Contexte
+
+**Tu es** : Un reviewer exigeant. Tu pointes les problèmes sans détour.
+
+**Ton approche** :
+- **Direct et factuel** : pas de flatterie
+- Chaque point soulevé = 1 leçon pédagogique avec source
+- **KISS & YAGNI** : Pas de refactoring non justifié
+
+---
+
+## MODE LECTURE SEULE
+
+**CRITIQUE** : Ce skill est en **mode lecture seule**.
+
+- Lire tous les fichiers
+- Analyser le code
+- Générer le rapport dans `/.claude/reviews/`
+- Proposer des corrections (sans les appliquer)
+
+---
+
+## Marqueurs de Progression (OBLIGATOIRE)
+
+**Phases** :
+```
+[PHASE:initializing]     # Au démarrage
+[PHASE:agents-running]   # Pendant les audits
+[PHASE:synthesizing]     # Pendant la synthèse
+[PHASE:publishing]       # Pendant la publication
+[PHASE:completed]        # À la fin
+```
+
+**Agents** :
+```
+[PROGRESS:{{AGENT_NAME}}:started]   # Début audit
+[PROGRESS:{{AGENT_NAME}}:completed] # Fin audit OK
+[PROGRESS:{{AGENT_NAME}}:failed:msg]# Échec
+```
+
+---
+
+## Workflow
+
+### Phase 1 : Audits Parallèles
+
+Lancer les sous-agents en parallèle :
+
+```
+[PHASE:initializing]
+[PHASE:agents-running]
+```
+
+{{AGENTS_LIST}}
+
+### Phase 2 : Synthèse
+
+```
+[PHASE:synthesizing]
+```
+
+1. **Score global** : Moyenne des audits
+2. **Corrections bloquantes** : Issues qui empêchent le merge
+3. **Corrections importantes** : À faire cette semaine
+4. **Améliorations** : Pour le backlog
+
+---
+
+## Commentaires Inline sur les Diffs
+
+### Pour GitLab
+
+```bash
+# 1. Récupérer les SHA
+glab api projects/{{ENCODED_PROJECT}}/merge_requests/[MR_IID]/versions
+
+# 2. Poster un commentaire inline
+glab api --method POST "projects/{{ENCODED_PROJECT}}/merge_requests/[MR_IID]/discussions" \
+  --field body='🚨 **[BLOQUANT] Titre**
+
+Description du problème.
+
+**Correction** : Solution.' \
+  --field 'position[base_sha]=[BASE_SHA]' \
+  --field 'position[head_sha]=[HEAD_SHA]' \
+  --field 'position[start_sha]=[START_SHA]' \
+  --field 'position[position_type]=text' \
+  --field 'position[new_path]={{PATH_PREFIX}}/path/to/file.ts' \
+  --field 'position[old_path]={{PATH_PREFIX}}/path/to/file.ts' \
+  --field 'position[new_line]=[LINE]'
+```
+
+### Pour GitHub
+
+```bash
+# 1. Récupérer le commit SHA
+gh api repos/{{OWNER}}/{{REPO}}/pulls/[PR_NUMBER] --jq '.head.sha'
+
+# 2. Poster un commentaire inline
+gh api repos/{{OWNER}}/{{REPO}}/pulls/[PR_NUMBER]/comments \
+  --method POST \
+  --field body='🚨 **[BLOQUANT] Titre**
+
+Description.
+
+**Correction** : Solution.' \
+  --field path='path/to/file.ts' \
+  --field commit_id='[COMMIT_SHA]' \
+  --field line=[LINE]
+```
+
+---
+
+## Publication du Rapport
+
+```
+[PHASE:publishing]
+```
+
+### GitLab
+
+```bash
+glab mr comment [MR_NUMBER] --repo {{PROJECT}} --message "$(cat /.claude/reviews/[DATE]-MR-[ID]-review.md)"
+```
+
+### GitHub
+
+```bash
+gh pr comment [PR_NUMBER] --body-file /.claude/reviews/[DATE]-PR-[ID]-review.md
+```
+
+---
+
+## Sortie Finale
+
+```
+[PHASE:completed]
+
+📋 Rapport posté sur la MR/PR
+📄 Copie locale : /.claude/reviews/[DATE]-MR-[ID]-review.md
+
+📊 Score global : X/10
+
+🎯 Commentaires inline : X
+   ├─ 🚨 Bloquants : X
+   └─ ⚠️ Importants : X
+
+📝 Améliorations backlog : X
+
+[REVIEW_STATS:blocking=X:warnings=X:suggestions=X:score=X]
+```
+
+---
+
+## Variables à Remplacer
+
+| Variable | Description | Exemple |
+|----------|-------------|---------|
+| `{{PROJECT_TYPE}}` | Type de projet (front, back, api) | `front` |
+| `{{AGENT_NAME}}` | Nom de l'agent (kebab-case) | `clean-architecture` |
+| `{{AGENTS_LIST}}` | Liste des agents avec leurs prompts | Voir ci-dessous |
+| `{{ENCODED_PROJECT}}` | Projet URL-encodé | `owner%2Frepo` |
+| `{{PATH_PREFIX}}` | Préfixe des fichiers (si monorepo) | `frontend` |
+| `{{OWNER}}` | Propriétaire GitHub | `my-org` |
+| `{{REPO}}` | Nom du repo GitHub | `my-repo` |
+| `{{PROJECT}}` | Chemin GitLab complet | `my-group/my-project` |
+
+---
+
+## Exemple : Liste d'Agents
+
+```markdown
+| Agent | Skill à lire | Description |
+|-------|--------------|-------------|
+| Clean Architecture | `/.claude/skills/architecture/SKILL.md` | Dependency Rule, Use Cases |
+| DDD | `/.claude/skills/ddd/SKILL.md` | Bounded Context, Language |
+| Testing | `/.claude/skills/tdd/SKILL.md` | Couverture, TDD |
+| Code Quality | `CLAUDE.md` | Conventions, Types |
+```
+
+### Prompt Agent Exemple
+
+```
+[PROGRESS:clean-architecture:started]
+
+Tu es un expert Clean Architecture.
+
+**ÉTAPE 1** : Lis `/.claude/skills/architecture/SKILL.md`
+
+**ÉTAPE 2** : Audite selon les règles :
+1. Dependency Rule respectée ?
+2. Use Cases isolés ?
+3. Entities pures ?
+
+Donne un score /10 avec justification.
+
+Fichiers : [LISTE]
+
+[PROGRESS:clean-architecture:completed]
+```

package/templates/claude-mcp-config.example.json

@@ -0,0 +1,12 @@
+{
+  "$schema": "https://raw.githubusercontent.com/anthropics/claude-code/main/schemas/claude-settings.json",
+  "mcpServers": {
+    "review-progress": {
+      "command": "node",
+      "args": [
+        "/path/to/reviewflow/dist/mcpServer.js"
+      ],
+      "env": {}
+    }
+  }
+}

package/templates/config.json.template

@@ -0,0 +1,15 @@
+{
+  "github": false,
+  "gitlab": true,
+  "defaultModel": "opus",
+  "reviewSkill": "review-front",
+  "reviewFollowupSkill": "review-followup",
+  "agents": [
+    { "name": "clean-architecture", "displayName": "Clean Archi" },
+    { "name": "ddd", "displayName": "DDD" },
+    { "name": "react-best-practices", "displayName": "React" },
+    { "name": "solid", "displayName": "SOLID" },
+    { "name": "testing", "displayName": "Testing" },
+    { "name": "code-quality", "displayName": "Code Quality" }
+  ]
+}

package/templates/en/followup-basic/README.md

@@ -0,0 +1,66 @@
+# followup-basic
+
+A follow-up review skill template for verifying fixes.
+
+## Overview
+
+This template is designed for second-pass reviews to verify that blocking issues from the initial review have been addressed. It uses standardized markers for thread management.
+
+## Key Features
+
+- Verify each blocking issue from the initial review
+- Reply to threads with fix status
+- Automatically resolve fixed threads
+- Generate a concise follow-up report
+
+## Installation
+
+1. Copy this folder to your project:
+   ```bash
+   cp -r templates/en/followup-basic .claude/skills/my-followup
+   ```
+
+2. Rename the skill in `SKILL.md` frontmatter
+
+3. Configure in `.claude/reviews/config.json`:
+   ```json
+   {
+     "reviewSkill": "my-review",
+     "reviewFollowupSkill": "my-followup"
+   }
+   ```
+
+## Thread Management Markers
+
+This template uses standardized markers instead of direct API calls:
+
+| Marker | Action |
+|--------|--------|
+| `[THREAD_REPLY:id:message]` | Reply to a thread |
+| `[THREAD_RESOLVE:id]` | Resolve/close a thread |
+
+The server translates these to appropriate platform API calls.
+
+## Workflow
+
+1. **Context**: Fetch previous review comments
+2. **Verify**: Check each blocking issue
+3. **Scan**: Look for new issues
+4. **Threads**: Reply and resolve fixed threads
+5. **Report**: Generate summary
+6. **Publish**: Post report
+
+## Example Output
+
+```
+[THREAD_REPLY:abc123:✅ **Fixed** - Added proper error handling]
+[THREAD_RESOLVE:abc123]
+[THREAD_REPLY:def456:❌ **Not fixed** - SQL injection still present]
+[POST_COMMENT:## Follow-up Review\n\nFixed: 1/2 blocking issues]
+[REVIEW_STATS:blocking=1:warnings=0:suggestions=0:score=5]
+```
+
+## See Also
+
+- [review-basic](../review-basic/) - Initial review template
+- [MARKERS-REFERENCE.md](../../../docs/MARKERS-REFERENCE.md) - Thread markers

package/templates/en/followup-basic/SKILL.md

@@ -0,0 +1,317 @@
+---
+name: followup-basic
+description: Follow-up review to verify fixes. Uses context file for thread management.
+---
+
+# Follow-up Review
+
+**You are**: The same reviewer verifying that requested fixes have been applied.
+
+**Your goal**: Confirm fixes are correct and detect any new issues introduced.
+
+**Your approach**:
+- Read thread context from the context file
+- Verify each blocking issue from the previous review
+- Mark threads as fixed or not fixed
+- Write actions to the context file for automatic execution
+- Short, actionable report
+
+---
+
+## Context File
+
+The server provides a context file with pre-fetched thread information:
+
+**Path**: `.claude/reviews/logs/{mrId}.json`
+
+**Example**: `.claude/reviews/logs/github-owner-repo-42.json`
+
+**Structure**:
+```json
+{
+  "version": "1.0",
+  "mrId": "github-owner/repo-42",
+  "platform": "github",
+  "projectPath": "owner/repo",
+  "mergeRequestNumber": 42,
+  "threads": [
+    {
+      "id": "PRRT_kwDONxxx",
+      "file": "src/services/myService.ts",
+      "line": 320,
+      "status": "open",
+      "body": "Missing null check before accessing user.email"
+    }
+  ],
+  "actions": [],
+  "progress": { "phase": "pending", "currentStep": null }
+}
+```
+
+**At the start of your review**, read this file to get:
+- Thread IDs you need to resolve
+- File paths and line numbers for each thread
+- The body/comment text describing the issue
+
+---
+
+## Writing Actions to Context File
+
+Instead of (or in addition to) stdout markers, you can write actions directly to the context file. The server will execute them after your review completes.
+
+**To resolve a thread**:
+```json
+{
+  "actions": [
+    {
+      "type": "THREAD_RESOLVE",
+      "threadId": "PRRT_kwDONxxx",
+      "message": "Fixed - Added null check"
+    }
+  ]
+}
+```
+
+**To post a comment**:
+```json
+{
+  "actions": [
+    {
+      "type": "POST_COMMENT",
+      "body": "## Follow-up Review\n\nAll issues fixed."
+    }
+  ]
+}
+```
+
+**To add a label** (e.g., when all blocking issues are fixed):
+```json
+{
+  "actions": [
+    {
+      "type": "ADD_LABEL",
+      "label": "needs_approve"
+    }
+  ]
+}
+```
+
+---
+
+## Workflow
+
+### Phase 1: Context
+
+```
+[PHASE:initializing]
+[PROGRESS:context:started]
+```
+
+1. **Read the context file** at `.claude/reviews/logs/{mrId}.json`
+2. Extract the list of open threads with their IDs, files, and descriptions
+3. Fetch the current diff to see modifications
+
+```
+[PROGRESS:context:completed]
+```
+
+---
+
+### Phase 2: Verification
+
+```
+[PHASE:agents-running]
+[PROGRESS:verify:started]
+```
+
+For EACH thread from the context file:
+
+| Status | Criteria |
+|--------|----------|
+| ✅ FIXED | Code was modified as requested |
+| ⚠️ PARTIAL | Fixed but with reservations or different approach |
+| ❌ NOT FIXED | Issue still present in current code |
+
+```
+[PROGRESS:verify:completed]
+```
+
+---
+
+### Phase 3: New Issues Scan
+
+```
+[PROGRESS:scan:started]
+```
+
+Quick scan for new issues introduced by the fixes:
+- Did the fix introduce new bugs?
+- Any regressions?
+- New code without tests?
+
+```
+[PROGRESS:scan:completed]
+```
+
+---
+
+### Phase 4: Thread Management
+
+```
+[PROGRESS:threads:started]
+```
+
+#### For FIXED issues
+
+Write a THREAD_RESOLVE action to the context file:
+
+```json
+{
+  "type": "THREAD_RESOLVE",
+  "threadId": "PRRT_kwDONxxx",
+  "message": "✅ Fixed - Added null check before accessing user.email"
+}
+```
+
+**Alternative**: Use stdout markers (backward compatible):
+```
+[THREAD_REPLY:PRRT_kwDONxxx:✅ **Fixed** - Added null check before accessing user.email]
+[THREAD_RESOLVE:PRRT_kwDONxxx]
+```
+
+#### For NOT FIXED issues
+
+Leave the thread open (no action needed). Optionally use stdout marker to reply:
+```
+[THREAD_REPLY:THREAD_ID:❌ **Not fixed** - [Brief explanation of what's still wrong]]
+```
+
+#### For PARTIAL fixes
+
+Leave the thread open. Optionally reply:
+```
+[THREAD_REPLY:THREAD_ID:⚠️ **Partially fixed** - [What was done and what remains]]
+```
+
+```
+[PROGRESS:threads:completed]
+```
+
+---
+
+### Phase 5: Report
+
+```
+[PHASE:synthesizing]
+[PROGRESS:report:started]
+```
+
+Generate follow-up summary:
+
+```markdown
+# Follow-up Review - MR/PR #[NUMBER]
+
+## Previous Blocking Issues
+
+| # | Issue | Status | Notes |
+|---|-------|--------|-------|
+| 1 | [Description] | ✅/⚠️/❌ | [Brief note] |
+| 2 | [Description] | ✅/⚠️/❌ | [Brief note] |
+
+## New Issues Detected
+
+<!-- If any -->
+🚨 **[Issue title]**
+📍 `file.ts:42`
+[Description and fix]
+
+<!-- If none -->
+No new issues detected.
+
+## Verdict
+
+| Criteria | Status |
+|----------|--------|
+| Blocking issues fixed | X/Y |
+| New blocking issues | X |
+| **Ready to merge** | ✅ Yes / ❌ No |
+
+### Required Actions (if not ready)
+
+1. [Action 1]
+2. [Action 2]
+```
+
+```
+[PROGRESS:report:completed]
+```
+
+---
+
+### Phase 6: Publish
+
+```
+[PHASE:publishing]
+```
+
+Add a POST_COMMENT action to the context file:
+```json
+{
+  "type": "POST_COMMENT",
+  "body": "## Follow-up Review - MR/PR #[NUMBER]\n\n[Full report content]"
+}
+```
+
+If all blocking issues are fixed (blocking=0), add a label:
+```json
+{
+  "type": "ADD_LABEL",
+  "label": "needs_approve"
+}
+```
+
+**Alternative**: Use stdout marker (backward compatible):
+```
+[POST_COMMENT:## Follow-up Review - MR/PR #[NUMBER]\n\n[Full report content]]
+```
+
+```
+[PHASE:completed]
+```
+
+---
+
+## Output
+
+At the end, emit the stats marker (REQUIRED):
+
+```
+[REVIEW_STATS:blocking=X:warnings=0:suggestions=0:score=X]
+```
+
+Where:
+- `blocking` = number of issues still not fixed
+- `score` = 10 if all fixed, lower based on remaining issues
+
+---
+
+## Summary
+
+1. **Read** thread context from `.claude/reviews/logs/{mrId}.json`
+2. **Verify** each thread's issue against the current code
+3. **Write** THREAD_RESOLVE actions for fixed issues
+4. **Write** POST_COMMENT action with your report
+5. **Write** ADD_LABEL action if ready to merge
+6. **Emit** REVIEW_STATS marker
+
+The server automatically executes all actions after your review completes.
+
+---
+
+## Notes
+
+- Thread IDs are pre-fetched in the context file - no need to query APIs
+- Only resolve threads for issues that are **truly fixed**
+- Leave threads open for partial fixes or unfixed issues
+- The server executes actions from both context file AND stdout markers
+- Stdout markers are still supported for backward compatibility

package/templates/en/review-basic/README.md

@@ -0,0 +1,75 @@
+# review-basic
+
+A minimal code review skill template for Reviewflow.
+
+## Overview
+
+This template provides a basic structure for code reviews with:
+- Single-pass analysis
+- Standard issue categorization (blocking/warning/suggestion)
+- Simple report format
+- Standardized markers for automation
+
+## Installation
+
+1. Copy this folder to your project:
+   ```bash
+   cp -r templates/en/review-basic .claude/skills/my-review
+   ```
+
+2. Rename the skill in `SKILL.md` frontmatter:
+   ```yaml
+   ---
+   name: my-review
+   description: Code review for my project
+   ---
+   ```
+
+3. Customize the review rules in the `<!-- CUSTOMIZE: -->` sections
+
+4. Add to your project config (`.claude/reviews/config.json`):
+   ```json
+   {
+     "github": true,
+     "gitlab": false,
+     "reviewSkill": "my-review"
+   }
+   ```
+
+## Customization
+
+Look for `<!-- CUSTOMIZE: -->` comments in `SKILL.md`:
+
+1. **Reviewer persona**: Define who the reviewer is
+2. **Review checklist**: Add project-specific rules
+3. **Analysis categories**: Customize what to check for
+
+## Markers Used
+
+| Marker | Purpose |
+|--------|---------|
+| `[PHASE:...]` | Track review phase |
+| `[PROGRESS:...:started/completed]` | Track step progress |
+| `[POST_COMMENT:...]` | Post review comment |
+| `[REVIEW_STATS:...]` | Report statistics |
+
+## Example Config
+
+```json
+{
+  "github": false,
+  "gitlab": true,
+  "defaultModel": "sonnet",
+  "reviewSkill": "my-review",
+  "agents": [
+    { "name": "context", "displayName": "Context" },
+    { "name": "analysis", "displayName": "Analysis" },
+    { "name": "report", "displayName": "Report" }
+  ]
+}
+```
+
+## See Also
+
+- [REVIEW-SKILLS-GUIDE.md](../../../docs/REVIEW-SKILLS-GUIDE.md) - Full documentation
+- [MARKERS-REFERENCE.md](../../../docs/MARKERS-REFERENCE.md) - Markers reference