| name | vtm-expert |
| description | Virtual Task Manager domain expert.
Knows about:
- Next, Context, Task, Start, Complete, Stats, List operations
Use when:
- User asks about task management operations
- Context needed before starting work
- Reviewing task status
- Managing task workflow
|
| trigger_phrases | ["what should i work on","next task","show my tasks","task status","vtm status","get context for task","start task","complete task","mark task done","task stats","list tasks"] |
VTM Expert Skill
What This Skill Does
Helps you manage your Virtual Task Manager workflow with smart command suggestions.
Auto-discovers when you mention task-related work and suggests the appropriate VTM command.
Available Commands
/vtm:next - Get next ready task to work on
/vtm:context - Generate minimal context for a specific task
/vtm:task - View details of a specific task
/vtm:start - Mark a task as in-progress
/vtm:complete - Mark a task as completed
/vtm:stats - Show VTM statistics and progress
/vtm:list - List all tasks with their status
When Claude Uses This
When you mention things like:
- "What should I work on?" → Suggests
/vtm:next
- "I need context for TASK-003" → Suggests
/vtm:context TASK-003
- "Show me task status" → Suggests
/vtm:stats or /vtm:list
- "Start working on TASK-005" → Suggests
/vtm:start TASK-005
- "Mark TASK-003 as done" → Suggests
/vtm:complete TASK-003
- "How many tasks are left?" → Suggests
/vtm:stats
VTM Workflow
The typical VTM workflow:
- Find work:
/vtm:next - Shows ready tasks (dependencies met)
- Get context:
/vtm:context TASK-XXX - Token-efficient task context
- Start task:
/vtm:start TASK-XXX - Mark as in-progress
- Implement: Use PROMPT 2 with TDD based on test_strategy
- Complete:
/vtm:complete TASK-XXX - Mark as done, unblocks dependents
- Repeat: Back to step 1
Token Efficiency
VTM achieves 99% token reduction by:
- Surgical access to specific tasks (not loading entire manifest)
- Dependency resolution built-in
- Two context modes: minimal (~2000 tokens) and compact (~500 tokens)
- No need to load related specs unless specifically needed
Best Practices
- Before starting: Run
/vtm:next to see what's ready
- Get context first: Always use
/vtm:context before implementing
- Follow TDD: Respect the task's
test_strategy field
- Track progress: Use
/vtm:stats to monitor overall progress
- Update status: Mark tasks in-progress and completed for accurate tracking
Integration
Works seamlessly with other Claude Code domains:
- Can be invoked manually:
/vtm:operation
- Auto-triggered by Claude based on conversation
- Integrates with git hooks for validation
- Supports team workflows via shared vtm.json
Customization
Edit trigger phrases in the frontmatter above to match your vocabulary.
Examples:
- Add project-specific terms: "backlog item", "sprint task"
- Add abbreviations: "ctx" for context, "nxt" for next
- Add workflow terms: "pick up task", "finish task"
Technical Details
Architecture:
- Commands wrap the vtm CLI (npm package)
- Requires vtm.json in project root
- Atomic writes with automatic stats recalculation
- Dependency validation built-in
Data Flow:
User phrase → Skill triggers → Command suggested →
User approves → vtm CLI executes → Stats auto-update
See Also
- Design spec:
.claude/designs/vtm.json
- Commands:
.claude/commands/vtm/
- Prompts:
prompts/1-generate-vtm.md, prompts/2-execute-task.md, prompts/3-add-feature.md