Initial commit
This commit is contained in:
105
commands/teds-checkpoint.md
Normal file
105
commands/teds-checkpoint.md
Normal file
@@ -0,0 +1,105 @@
|
||||
---
|
||||
description: Create a checkpoint in the current TEDS task
|
||||
---
|
||||
|
||||
# Create TEDS Checkpoint
|
||||
|
||||
Create an immediate checkpoint in the currently active TEDS task.
|
||||
|
||||
## Usage
|
||||
|
||||
```bash
|
||||
/teds-checkpoint [optional-task-id]
|
||||
```
|
||||
|
||||
## Examples
|
||||
|
||||
```bash
|
||||
# Create checkpoint in current task context
|
||||
/teds-checkpoint
|
||||
|
||||
# Create checkpoint for specific task
|
||||
/teds-checkpoint 20250116-1430-refactor-auth
|
||||
```
|
||||
|
||||
## What is a Checkpoint?
|
||||
|
||||
A checkpoint is a **safe pause point** that captures:
|
||||
- Current phase and progress percentage
|
||||
- Summary of what has been accomplished
|
||||
- What remains to be done
|
||||
- Timestamp for recovery reference
|
||||
|
||||
Think of it as a "save game" point—you can safely stop work and resume from here later.
|
||||
|
||||
## What This Does
|
||||
|
||||
The agent will:
|
||||
1. Review recent work from `execution_log.md`
|
||||
2. Assess current progress
|
||||
3. Add checkpoint entry to `execution_log.md`:
|
||||
```markdown
|
||||
### [HH:MM] - CHECKPOINT
|
||||
- Phase: [current phase]
|
||||
- Progress: [X%]
|
||||
- Summary: [accomplishments]
|
||||
- Next: [what remains]
|
||||
```
|
||||
4. Update `last_checkpoint` in `status.yaml`
|
||||
5. Confirm: "Checkpoint created. Safe to pause here."
|
||||
|
||||
## Automatic Checkpoints
|
||||
|
||||
The executor agent automatically creates checkpoints:
|
||||
- Every 30+ minutes of active work
|
||||
- At completion of major milestones
|
||||
- When transitioning between phases
|
||||
|
||||
Manual checkpoints are useful when:
|
||||
- You want to pause before the automatic interval
|
||||
- You've reached a logical stopping point
|
||||
- You're about to try something risky
|
||||
- You're switching focus to another task
|
||||
|
||||
## Checkpoint Best Practices
|
||||
|
||||
**Good times to checkpoint**:
|
||||
- ✅ Just completed a significant feature
|
||||
- ✅ About to refactor or make large changes
|
||||
- ✅ End of work session
|
||||
- ✅ Before switching context
|
||||
|
||||
**Don't checkpoint**:
|
||||
- ❌ In the middle of an incomplete action
|
||||
- ❌ When build/tests are failing
|
||||
- ❌ When blocked and unsure how to proceed (mark as blocked instead)
|
||||
|
||||
## Viewing Checkpoints
|
||||
|
||||
To see all checkpoints in a task:
|
||||
```bash
|
||||
# View full execution log
|
||||
cat workspace/active_tasks/[task-id]/execution_log.md | grep "CHECKPOINT"
|
||||
```
|
||||
|
||||
Or use:
|
||||
```bash
|
||||
/teds-status [task-id]
|
||||
```
|
||||
This shows the last checkpoint time.
|
||||
|
||||
## Recovery from Checkpoint
|
||||
|
||||
When you continue a task with `/teds-continue`, the agent automatically:
|
||||
1. Finds the most recent checkpoint
|
||||
2. Reviews what was done
|
||||
3. Identifies the next action
|
||||
4. Resumes from there
|
||||
|
||||
No special recovery command needed—it's built into the continue process.
|
||||
|
||||
## Related Commands
|
||||
|
||||
- `/teds-continue [task-id]` - Resume from checkpoint
|
||||
- `/teds-status` - View last checkpoint time
|
||||
- `/teds-complete [task-id]` - Finish task
|
||||
Reference in New Issue
Block a user