@ranger-testing/ranger-cli 1.1.3 → 1.1.5

package/build/skills/ranger/feedback.md

@@ -0,0 +1,64 @@
+# Addressing Reviewer Feedback
+
+When a reviewer leaves comments on checklist items, a new version (v2) of the item is created to incorporate the feedback. The reviewer's comments were left on the **previous version** (v1), but they are the reason this v2 item exists. Use this workflow to understand what the reviewer wants, fix the code, and re-verify.
+
+v2+ items inherit comments from their parent version. Run `get-feedback` to see them.
+
+## When This Applies
+
+- Items show as v2/v3 with `[has reviewer feedback]` tag
+- `ranger feature resume` prints a warning about unaddressed comments
+
+## Step 1: Read the Feedback
+
+```bash
+ranger feature get-feedback
+```
+
+This shows ALL reviewer comments across all items, including:
+- The specific comment content and who wrote it
+- The previous version description (what changed from v1 to v2)
+- The expected flow from prior verification (canonical flow)
+
+### Example Output
+
+```
+📋 Feedback for: User Authentication (feat_abc123)
+
+Item 1: "User can log in" (v2) — 2 unaddressed comments (from previous version)
+  Previous version: "User can log in with email"
+  This item was created to address the following reviewer feedback:
+  💬 Jane (Feb 5): "Password field should show/hide toggle"
+  💬 Bob (Feb 5): "Error message not visible on failed login"
+  Expected flow:
+    1. Navigate to /login
+    2. Enter email and password
+    3. Click Submit
+    4. Verify dashboard appears
+
+Item 2: "Dashboard loads" — 0 comments
+  ✅ No feedback to address
+```
+
+## Step 2: Fix the Code
+
+For each unaddressed comment:
+1. Read the comment carefully — what is the reviewer asking for?
+2. Look at the "previous version" description to understand what changed
+3. If a canonical flow is provided, that's the expected user journey to verify against
+4. Make the code changes that address each concern
+
+## Step 3: Re-verify
+
+```bash
+ranger verify-feature --item <N>
+```
+
+The verification agent **automatically receives the reviewer comments** in its prompt. It will specifically check that each reviewer concern was addressed in the current implementation.
+
+## Key Points
+
+- Always run `get-feedback` **before** making code changes — understand what the reviewer wants first
+- v2+ items inherit their parent's comments — those comments are the reason this version exists
+- The verify-feature command auto-injects feedback into the browser agent's prompt, so you don't need to manually include comments in your `--task` description
+- Address ALL unaddressed comments before re-verifying — partial fixes may result in another review round

package/build/skills/ranger/start.md

@@ -45,6 +45,16 @@ This displays:
 - Git context (repo, branch)
 - Checklist with item statuses
 
+## Check for Reviewer Feedback
+
+If any checklist items show comment badges (e.g., `[2 comments]`) or are at v2+, reviewer feedback needs to be addressed:
+
+```bash
+ranger feature get-feedback
+```
+
+This shows the actual comment content, who wrote it, and the previous version description. **Read [feedback.md](./feedback.md) for the full feedback workflow.**
+
 ## Add Checklist Items
 
 If you need to add new work to an existing feature:
@@ -88,7 +98,13 @@ ranger feature list
      │                                ranger feature show
      │                                         │
      │                                         ▼
-     │                                Continue working
+     │                                Items have comments?
+     │                                    │          │
+     │                                    YES        NO
+     │                                    │          │
+     │                                    ▼          ▼
+     │                           get-feedback   Continue working
+     │                           fix + verify
      │
      └── None exist? ──▶ See create.md
 ```

package/build/skills/ranger/verify.md

@@ -22,10 +22,20 @@ ranger feature resume <id>         # Resume a specific feature
 ## The Verification Flow
 
 1. **Select checklist item** - CLI prompts which item this verifies
-2. **Run browser verification** - Agent executes the task in a real browser
-3. **Evaluate results** - Agent determines if the checklist item is satisfied
-4. **Update status** - Item is marked verified, partial, blocked, or failed
-5. **Link evidence** - Session trace is attached to the item
+2. **Fetch reviewer feedback** - If the item has unaddressed comments or a parent item, reviewer comments are automatically injected into the verification prompt
+3. **Run browser verification** - Agent executes the task in a real browser
+4. **Evaluate results** - Agent determines if the checklist item is satisfied (including whether reviewer concerns were addressed)
+5. **Update status** - Item is marked verified, partial, blocked, or failed
+6. **Link evidence** - Session trace is attached to the item
+
+### Reviewer Feedback Auto-Injection
+
+When verifying items that have reviewer comments (v2+ items or items with unaddressed comments), the verification agent automatically receives a **"Reviewer Feedback to Address"** section in its prompt. This includes:
+- Each reviewer comment with author and date
+- The previous version's description
+- The canonical flow from prior verification (if available)
+
+You do NOT need to manually include reviewer feedback in your `--task` description — it's handled automatically. Just make sure you've addressed the feedback in your code before verifying.
 
 ## Options
 
@@ -151,7 +161,7 @@ Check progress:
 ranger feature show
 ```
 
-If all non-canceled items are verified, the feature auto-completes:
+If all non-closed items are verified, the feature auto-completes:
 
 ```
 ✅ User Authentication (feat_abc123)

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@ranger-testing/ranger-cli",
-    "version": "1.1.3",
+    "version": "1.1.5",
     "type": "module",
     "bin": {
         "ranger": "./build/cli.js"
@@ -20,7 +20,6 @@
         "@anthropic-ai/claude-agent-sdk": "^0.1.0",
         "dotenv": "^16.4.5",
         "inquirer": "^9.2.12",
-        "keytar": "^7.9.0",
         "playwright": "^1.57.0",
         "yargs": "^17.7.2",
         "zod": "^3.23.8"