Complete conversation lifecycle management - start (understanding & alignment), capture (mid-conversation context), and exit (debrief & improvement). Use automatically at conversation boundaries and checkpoints.
This skill inherits all available tools. When active, it can use any tool Claude has access to.
Manages the full conversation lifecycle: start, capture, and exit.
| Phase | When | Purpose |
|---|---|---|
| Start | Beginning of conversation | Establish understanding and alignment |
| Capture | Mid-conversation | Save context for continuity |
| Exit | End of conversation | Debrief, feedback, and improvement |
Establishes understanding and alignment before diving into execution.
Actions:
Research Priority Pattern:
I understand you want to [specific goal].
This involves:
- [Scope item 1]
- [Scope item 2]
Here's my approach:
1. [Step 1]
2. [Step 2]
3. [Step 3]
I'll use:
- [Relevant agent/pattern]
- [Another relevant pattern]
Do NOT proceed until user confirms or corrects the plan.
For simple tasks:
Got it - you want [X]. I'll [approach]. Sound good?
For complex tasks:
Let me make sure I understand:
**Goal**: [What you want to achieve]
**Scope**: [What's included/excluded]
**Approach**: [How I'll tackle it]
**Output**: [What you'll get]
Anything to adjust before I start?
Skip when:
Never skip when:
Saves context mid-conversation for continuity and handoff.
| Element | Why It Matters |
|---|---|
| Current state | Where we are right now |
| Decisions made | What was decided and why |
| Files changed | What was created/modified |
| Next steps | What comes next |
---
**Checkpoint**: [Brief description]
**Status**: [Current state of work]
**Just completed**: [What was done]
**Next**: [What's queued]
---
Create convo-captures/YYYY-MM-DD-HHMM-topic.md:
# Context Capture: [Topic]
**Date**: YYYY-MM-DD HH:MM
**Status**: In Progress / Paused / Blocked
## Current State
[Where we are in the work]
## Decisions Made
| Decision | Rationale | Alternatives Considered |
|----------|-----------|------------------------|
| [Choice] | [Why] | [Other options] |
## Files Changed
**Created**: [list]
**Modified**: [list]
## Next Steps
1. [ ] [Immediate next step]
2. [ ] [Following step]
## To Resume
[Instructions for picking this back up]
Wraps up conversations with debrief, feedback, and knowledge capture.
Let me review our conversation:
**Original goal**: [State the goal]
**What we accomplished**:
- [Outcome 1]
- [Outcome 2]
**Files created/modified**:
- `path/file.md` - [Brief description]
**Tasks Remaining**:
For this effort:
- [ ] [Remaining task 1]
- [ ] [Remaining task 2]
**Your feedback**:
1. What worked well in this conversation?
2. What would you have liked to see improved?
3. Were there moments where I was unclear or off-track?
After receiving feedback:
**Based on our conversation, I propose these changes**:
1. [File path]: [Specific change]
2. [File path]: [Specific change]
Should I make these changes?
File: convo-exits/YYYY-MM-DD-HHMM-topic.md
# Conversation Exit: [Topic]
**Date**: YYYY-MM-DD HH:MM
**Repo**: [which repo(s) were involved]
## Original Goal
[What we set out to do]
## Outcomes
**Files Created**: [list]
**Files Modified**: [list]
**Decisions Made**: [list]
## User Feedback
**What worked well**: [feedback]
**What could improve**: [feedback]
## Improvements Made
- [ ] [Change 1] - [Status]
## Next Steps
**Immediate**: [list]
**Future**: [list]
If brain repo has personal/activity-log.md AND significant work was done:
## YYYY-MM-DD HH:MM - [Topic]
**What I accomplished**:
- [Accomplishment 1]
- [Accomplishment 2]
**Repos affected**: [list]
**Related**: `convo-exits/YYYY-MM-DD-HHMM-topic.md`
| Wrong | Right |
|---|---|
| Immediately starting work | State understanding first, get confirmation |
| Acting on assumptions | Explicitly state assumptions, ask for confirmation |
| No record of decisions | Capture decisions with rationale |
| Context dump without resume instructions | Clear "To continue, do X" instructions |
| Skipping debrief | Always offer exit protocol for significant work |