# Linear Testing Plan - One Thing at a Time
## Understanding from Existing Files
### Plugin Structure (Discovered):
- **Agents**: `{plugin}/agents/{name}.md` with frontmatter:
- `name`: Agent identifier
- `description`: What the agent does
- `tools`: Comma-separated list of tools
- `model`: Model to use (sonnet, opus, etc.)
- `color`: UI color (red, yellow, green, etc.)
- **Skills**: `{plugin}/skills/{name}/SKILL.md` with frontmatter:
- `name`: Skill identifier
- `description`: What the skill does
- `license`: License info
- **Commands**: `{plugin}/commands/{name}.md` with frontmatter:
- `description`: Command description
- `argument-hint`: Optional argument hint
- **Hooks**: `{plugin}/hooks/hooks.json` with:
- `description`: Hook description
- `hooks`: Object mapping hook types to handlers
### Settings Structure (Discovered):
- Global: `~/.claude/settings.json`
- Local: `~/.claude/settings.local.json`
- Project: `.claude/` directory (needs testing)
### MCP Configuration (Discovered):
- `--mcp-config` accepts JSON files or strings
- `claude mcp` commands for management
- Per-instance configs via flags
## Phase 1: Research Complete ✅
**Status:** Based on existing files and CLI help, we have enough to proceed with testing.
## Phase 2: Testing - ONE THING AT A TIME
### Test 1: Query Session ID ✅ READY
**Goal:** Extract session ID from Claude output
**Test:**
```bash
claude --print --model sonnet --output-format json "test" | jq -r '.session_id'
```
**Expected:** UUID string
**Success Criteria:** Can extract session ID programmatically
---
### Test 2: Run 2 Instances Simultaneously ✅ READY
**Goal:** Verify multiple instances can run at once
**Test:**
```bash
# Terminal 1
claude --print --model sonnet "Instance 1" &
# Terminal 2
claude --print --model opus "Instance 2" &
```
**Expected:** Both complete successfully
**Success Criteria:** Both instances complete without errors
---
### Test 3: Different MCP Config Per Instance ✅ READY
**Goal:** Each instance has different MCP servers
**Test:**
```bash
# Instance 1: Only imagine MCP
claude --print --mcp-config ./claude_config.json "test" &
# Instance 2: Only context7 MCP (default)
claude --print "test" &
```
**Expected:** Each instance sees different tools
**Success Criteria:** Tool lists differ between instances
---
### Test 4: Different Tools Per Instance ✅ READY
**Goal:** Each instance has different tool sets
**Test:**
```bash
# Instance 1: Only Read tool
claude --print --tools "Read" "test" &
# Instance 2: Only Write tool
claude --print --tools "Write" "test" &
```
**Expected:** Each instance sees different tools
**Success Criteria:** Tool lists differ
---
### Test 5: Different Agents Per Instance ✅ READY
**Goal:** Each instance has different agents
**Test:**
```bash
# Instance 1: Reviewer agent
claude --print --agents '{"reviewer": {"description": "...", "prompt": "..."}}' "test" &
# Instance 2: Planner agent
claude --print --agents '{"planner": {"description": "...", "prompt": "..."}}' "test" &
```
**Expected:** Each instance has different agents available
**Success Criteria:** Agents differ between instances
---
### Test 6: Different Settings Per Instance ✅ READY
**Goal:** Each instance uses different settings
**Test:**
```bash
# Instance 1: Custom settings file
claude --print --settings ./settings1.json "test" &
# Instance 2: Different settings file
claude --print --settings ./settings2.json "test" &
```
**Expected:** Settings applied per instance
**Success Criteria:** Behavior differs based on settings
---
### Test 7: Project-Specific Configuration ✅ READY
**Goal:** Claude in project folder uses project configs
**Test:**
```bash
# Create project with .claude directory
mkdir -p test-project/.claude
# Add config files
cd test-project
claude --print "test"
```
**Expected:** Project configs are used
**Success Criteria:** Project-specific behavior
---
### Test 8: Expert Instance Configuration ✅ READY
**Goal:** Create specialized expert instance
**Test:**
```bash
# Code reviewer expert
claude --print \
--model sonnet \
--agents '{"reviewer": {"description": "Code reviewer", "prompt": "You are an expert code reviewer..."}}' \
--tools "Read,Grep" \
--system-prompt "You are a code review expert" \
"Review this code: ..."
```
**Expected:** Expert behaves as specialized reviewer
**Success Criteria:** Expert shows specialized behavior
---
### Test 9: Plugin System Understanding ✅ READY
**Goal:** Understand how plugins work
**Test:**
```bash
# List plugins
claude plugin list
# Install plugin
claude plugin install <plugin>
# Use plugin
claude --plugin-dir <path> "test"
```
**Expected:** Plugins can be installed and used
**Success Criteria:** Plugins work as expected
---
### Test 10: Skills Library ✅ READY
**Goal:** Create and use custom skills
**Test:**
```bash
# Create skill
# Use skill via plugin
```
**Expected:** Skills can be created and used
**Success Criteria:** Skills work as expected
---
## Execution Order
**CRITICAL:** Test ONE thing at a time. Don't proceed to next test until current test is:
1. ✅ Understood
2. ✅ Tested
3. ✅ Documented
4. ✅ Verified working
## Current Status
**Next Test:** Test 1 - Query Session ID
**Status:** Ready to execute
