ktg-plugin-marketplace/plugins/linkedin-thought-leadership/docs/integration-test-guide.md

Integration Test Guide: LinkedIn Thought Leadership Plugin

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

Prerequisites

Before testing, ensure:

• ~/.claude/linkedin-thought-leadership.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 thought-leadership-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-thought-leadership.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-thought-leadership.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, velocity score, anomaly detection, and recommended actions 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, significance assessment (20% rule), 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