linkedin-studio/docs/integration-test-guide.md

# Integration Test Guide: LinkedIn Studio Plugin

Manual integration testing scenarios for commands, agents, and hooks in the plugin.

## Prerequisites

Before testing, ensure:
- [ ] `~/.claude/linkedin-studio.local.md` exists (create from `config/state-file.template.md`)
- [ ] Voice samples exist in `assets/voice-samples/authentic-voice-samples.md`
- [ ] Quality scorecard exists at `assets/checklists/quality-scorecard.md`
- [ ] Plugin is installed: appears in Claude Code's skill/command list

## /linkedin:pipeline — End-to-End Tests

### Test 1: Full Pipeline — Idea to Post

**Goal:** Execute the complete 8-step pipeline from ideation to publish-ready post.

**Steps:**
1. Run `/linkedin:pipeline`
2. Verify Step 0 loads: state file read, status displayed (posts/week, streak)
3. Choose "Generate ideas for me" when prompted
4. Verify 3 topic suggestions appear, drawn from `content-angles.md`
5. Select a topic → verify angle selection (2-3 options)
6. Choose format → verify draft follows structure (hook/context/insight/implication/CTA)
7. Verify optimization checks run:
   - Hook: 110-140 chars
   - Total: 1,200-1,800 chars
   - No external links in body
   - No corporate buzzwords
8. Verify scheduling recommendation mentions CET times
9. Verify 5x5x5 guidance is provided
10. Verify copy-paste ready output with character count and hashtags
11. Verify first-hour monitoring plan is shown
12. Verify 48-hour check-in reminder appears

**Expected outcome:** A complete, publish-ready post with all quality checks passed.

**Hooks that fire:**
- `SessionStart` → loads state
- `UserPromptSubmit` → injects context
- `PreToolUse (Write)` → quality gate + voice guardian (if draft is written to file)
- `PostToolUse (Write)` → alternative hooks + posting time suggestion
- `Stop` → state update + pre-publish reminders

### Test 2: Pipeline with Existing Topic

**Goal:** User provides their own topic, skipping ideation.

**Steps:**
1. Run `/linkedin:pipeline`
2. Choose "I have an idea already"
3. Provide topic: "Why AI agents will replace workflows in 2026"
4. Verify the topic is used directly (no override)
5. Verify angle suggestions are relevant to the provided topic
6. Complete the remaining steps

**Expected outcome:** Post is created on the user's topic, not a generated one.

### Test 3: Pipeline with State File Missing

**Goal:** Graceful handling when state file doesn't exist.

**Steps:**
1. Temporarily rename `~/.claude/linkedin-studio.local.md`
2. Run `/linkedin:pipeline`
3. Verify: no crash, reasonable fallback (e.g., "No posting data found. Starting fresh.")
4. Complete the pipeline
5. Verify: state file is created after pipeline completes

**Expected outcome:** Pipeline works without state file, creates one at the end.

### Test 4: Pipeline — Draft Save Option

**Goal:** Verify "Save as draft for later" works.

**Steps:**
1. Run `/linkedin:pipeline`
2. Create a post
3. At scheduling step, choose "Save as draft for later"
4. Verify: no posting reminders (5x5x5, first-hour) are shown for drafts
5. Verify: state file is NOT updated with post date (it's a draft, not published)

**Expected outcome:** Draft is saved without publishing-related actions.

---

## /linkedin:batch — End-to-End Tests

### Test 5: Full Batch — 3 Posts from One Theme

**Goal:** Create 3 posts from a single theme with varying angles and formats.

**Steps:**
1. Run `/linkedin:batch`
2. Verify Step 0 loads: state file, check for existing weekly plan
3. Choose "One main theme"
4. Provide theme: "The future of AI in public sector"
5. Verify batch plan shows 3 posts with:
   - Different angles (not repetitive)
   - Mixed formats (not all the same)
   - Different target days
6. Approve the plan
7. Verify each post:
   - Follows structure (hook 110-140 chars, 1,200-1,800 total)
   - Has unique angle
   - Quick quality check passes
8. Verify posts are saved to `assets/drafts/week-[WXX]/`
9. Verify filenames follow pattern: `[day]-[topic-slug].md`
10. Verify YAML frontmatter in each file (planned_date, pillar, angle, format, status)
11. Verify summary shows content mix and pillar coverage
12. Approve all drafts
13. Verify posting schedule with recommended times

**Expected outcome:** 3 distinct posts saved in correct directory with proper metadata.

### Test 6: Batch — Content Pillar Mode

**Goal:** Batch using existing content pillar.

**Steps:**
1. Run `/linkedin:batch`
2. Choose "Content pillar"
3. Select from user's defined pillars in skill file
4. Verify posts are created around that pillar
5. Verify angle variety (not same perspective repeated)

**Expected outcome:** All posts align with chosen pillar but explore different angles.

### Test 7: Batch — Revision Flow

**Goal:** Verify post revision during batch creation.

**Steps:**
1. Run `/linkedin:batch` and create 3 posts
2. At review step, choose "Revise a specific post"
3. Ask for post #2 to be revised (e.g., "Make the hook more provocative")
4. Verify: only post #2 is changed, others remain intact
5. Verify: summary updates to reflect the revised post

**Expected outcome:** Individual post revision works without affecting other batch posts.

### Test 8: Batch — Drafts Directory Creation

**Goal:** Verify `assets/drafts/` directory is created when it doesn't exist.

**Steps:**
1. Ensure `assets/drafts/` does not exist
2. Run `/linkedin:batch` and complete the workflow
3. Verify: `assets/drafts/week-[WXX]/` directory is created
4. Verify: all posts are saved correctly

**Expected outcome:** Directory is created automatically, posts are saved.

---

## Cross-Command Integration Tests

### Test 9: Pipeline After Batch

**Goal:** Pipeline uses batch-created drafts.

**Steps:**
1. First run `/linkedin:batch` to create 3 drafts
2. Then run `/linkedin:pipeline`
3. At ideation, choose "Use a planned topic"
4. Verify: pipeline picks up a draft from the batch
5. Complete pipeline with the batch draft
6. Verify: state file is updated after publishing

**Expected outcome:** Pipeline can consume batch-created drafts seamlessly.

### Test 10: Batch Respects Weekly State

**Goal:** Batch adjusts recommendations based on current posting state.

**Steps:**
1. Set state file to show 2 posts already published this week
2. Run `/linkedin:batch` with goal of 3 posts/week
3. Verify: batch suggests creating only 1 post (3 - 2 = 1 remaining)
4. Or if configurable, verify batch mentions current progress

**Expected outcome:** Batch is aware of weekly posting status.

---

## Hook Integration Tests

### Test 11: Quality Gate Fires on Post Draft

**Goal:** Verify PreToolUse quality gate hook catches issues.

**Steps:**
1. During pipeline or batch, intentionally create a post with:
   - Hook over 140 chars
   - External link in body
   - Corporate buzzword ("leverage")
2. Verify: quality gate flags ALL issues
3. Verify: issues are described specifically (not generic warnings)

**Expected outcome:** Quality gate catches all three violations with specific feedback.

### Test 12: Voice Guardian Detects AI Patterns

**Goal:** Verify voice guardian hook catches AI-sounding content.

**Steps:**
1. During pipeline, create a post that starts with "In today's rapidly evolving landscape..."
2. Verify: voice guardian flags the AI pattern
3. Verify: specific rewrite suggestions are provided
4. Verify: voice samples are referenced for comparison (if they exist)

**Expected outcome:** Voice guardian identifies AI patterns and suggests authentic alternatives.

### Test 13: Stop Hook Updates State

**Goal:** Verify session-end state update works correctly.

**Steps:**
1. Run `/linkedin:pipeline` and create a post
2. Note the topic and hook
3. End the session (or let Stop hook fire)
4. Read `~/.claude/linkedin-studio.local.md`
5. Verify:
   - `last_post_date` = today
   - `last_post_topic` = the topic used
   - `posts_this_week` incremented
   - `current_streak` updated correctly
   - Recent Posts section has new entry

**Expected outcome:** State file accurately reflects the session's output.

### Test 14: PostToolUse Generates Alternative Hooks

**Goal:** Verify post-creation automation fires.

**Steps:**
1. During pipeline or batch, write a post draft
2. Verify: 3 alternative hooks are generated
3. Verify: each alternative has character count shown
4. Verify: optimal posting time is suggested
5. Verify: 5x5x5 reminder appears

**Expected outcome:** Post-creation automation provides actionable suggestions.

---

## Agent Tests

### Test 15: Post-Feedback Monitor — Basic Monitoring
**Command:** Trigger `post-feedback-monitor` agent
**Steps:**
1. Say "How is my latest post doing?"
2. Agent should load algorithm-signals-reference and engagement-frameworks
3. Agent should ask which post to monitor
4. Provide sample metrics: 500 impressions, 15 reactions, 3 comments, 1 repost
5. Agent should identify the current phase and provide benchmarks
**Expected:** Structured output with metrics snapshot, a baseline reading (above/within/below band, or an explicit "no verdict — N post(s), 5 required" refusal), anomaly detection, and recommended actions. No composite score.
**Validates:** Agent file loads correctly, context loading works, output format matches spec

### Test 16: Post-Feedback Monitor — Anomaly Detection
**Command:** Trigger `post-feedback-monitor` agent
**Steps:**
1. Say "My post has 2000 impressions but only 5 reactions"
2. Agent should detect "Impression-Engagement Gap" anomaly
3. Agent should provide specific intervention recommendations
**Expected:** Anomaly correctly identified with cause analysis and action plan
**Validates:** Anomaly detection framework, intervention playbook

### Test 17: Post-Feedback Monitor — Golden Hour
**Command:** Trigger `post-feedback-monitor` agent
**Steps:**
1. Say "I just posted 30 minutes ago, what should I do?"
2. Agent should activate Golden Hour protocol
3. Agent should provide time-sensitive action items
**Expected:** Golden Hour specific advice (reply within 5 min, DM connections, first comment strategy)
**Validates:** Phase detection, time-sensitive interventions

---

## Command Tests

### Test 18: A/B Test — Design New Test
**Command:** `/linkedin:ab-test`
**Steps:**
1. Run the command
2. Select "Design a new A/B test"
3. Choose "Hook/Opening line" as the variable
4. Follow the guided workflow
**Expected:** Complete test plan with hypothesis, variants, execution schedule, success criteria
**Validates:** Command loads, AskUserQuestion flow works, reference file loads, test plan file created

### Test 19: A/B Test — Analyze Results
**Command:** `/linkedin:ab-test`
**Steps:**
1. First create a test plan (Test 18) and manually create a test file with sample data
2. Run `/linkedin:ab-test` and select "Analyze test results"
3. Select the test to analyze
**Expected:** Results comparison table, directional assessment (≥20% gap), verdict, recommended next steps
**Validates:** File scanning, data analysis, result formatting

### Test 20: Enhanced Report — Trends & Alerts
**Command:** `/linkedin:report`
**Steps:**
1. Ensure at least 4 weeks of imported data exists
2. Run `/linkedin:report` for the current week
3. Verify trend analysis section appears after main report
4. Verify alert detection section appears
**Expected:** 4-week trend table, trend interpretation, performance alerts, algorithm alerts
**Validates:** Trend CLI integration, alert thresholds, formatting

### Test 21: Enhanced Import — Anomaly Detection
**Command:** `/linkedin:import`
**Steps:**
1. Ensure baseline data exists (previous imports)
2. Import a new CSV export
3. After import, verify anomaly detection runs
**Expected:** Breakout posts flagged, patterns detected, intelligent next steps offered
**Validates:** Anomaly detection rules, baseline comparison, conditional suggestions

### Test 22: Enhanced Report — Markdown Export
**Command:** `/linkedin:report`
**Steps:**
1. Run `/linkedin:report` for any week with data
2. Select "Export as Markdown" from options
3. Verify file is saved to `assets/analytics/weekly-reports/YYYY-WXX-report.md`
**Expected:** Clean markdown file with all sections (metrics, trends, alerts, top performers, recommendations)
**Validates:** Export template, file creation, gitignore compliance

---

## Cross-Command Integration Tests

### Test 23: Router — New Commands Accessible
**Command:** `/linkedin`
**Steps:**
1. Run `/linkedin`
2. Verify A/B test appears in command menu
3. Verify post-feedback-monitor appears in agent suggestions
4. Say "I want to A/B test my hooks" — should route to `/linkedin:ab-test`
5. Say "How is my post doing?" — should route to `post-feedback-monitor`
**Expected:** All new commands and agents are accessible through the router
**Validates:** Router updates, intent matching

### Test 24: Collaboration — Multi-Author Workflow
**Command:** `/linkedin:collab`
**Steps:**
1. Run `/linkedin:collab` and complete readiness check
2. Navigate to multi-author content coordination section
3. Verify co-creation workflow templates are available
4. Verify collaboration tracking section exists
**Expected:** Multi-author workflow with 5 phases, shared draft guidelines, collaboration pipeline board
**Validates:** New collab command sections (Step 7 and Step 8)

---

## Known Limitations

1. **No automated testing:** These commands are conversational — they require human interaction at AskUserQuestion steps. Testing must be manual.

2. **State file format:** State file uses YAML frontmatter. Any malformed YAML will cause parsing issues. Always validate format after manual edits.

3. **Draft directory:** `assets/drafts/` and `assets/plans/` are created at runtime. They don't exist in the base plugin directory and won't appear until first use.

4. **Hook ordering:** PreToolUse has two hooks (quality gate + voice guardian). Both fire on every Write/Edit of content files. If one blocks, the user must fix the issue before proceeding.

5. **Content vs. config detection:** All prompt-based hooks include logic to skip non-content files. This relies on heuristic pattern matching (checking for `.local.md`, `.json`, script extensions, etc.). Edge cases may exist.

6. **Agent testing:** Agents (Tests 15-17) are triggered conversationally, not via slash commands. They require natural language input and cannot be invoked deterministically. Test by using the trigger phrases documented in the agent frontmatter.

7. **Structure validation:** Use `scripts/test-runner.sh` to validate file existence, frontmatter format, and router completeness. This is automated and complements the manual integration tests above.

## Test Results Log

Record results here when tests are executed:

| Test | Date | Result | Notes |
|------|------|--------|-------|
| 1    |      |        |       |
| 2    |      |        |       |
| 3    |      |        |       |
| 4    |      |        |       |
| 5    |      |        |       |
| 6    |      |        |       |
| 7    |      |        |       |
| 8    |      |        |       |
| 9    |      |        |       |
| 10   |      |        |       |
| 11   |      |        |       |
| 12   |      |        |       |
| 13   |      |        |       |
| 14   |      |        |       |
| 15   |      |        |       |
| 16   |      |        |       |
| 17   |      |        |       |
| 18   |      |        |       |
| 19   |      |        |       |
| 20   |      |        |       |
| 21   |      |        |       |
| 22   |      |        |       |
| 23   |      |        |       |
| 24   |      |        |       |