saving-progress

Saving Progress

Manages work context through saves, restores, and recovery from interruptions. Acts as external working memory for maintaining state across task switches and disruptions.

When to Use This Skill

Use this skill when detecting:

• Save requests: "save where I am", "checkpoint this", "need to switch tasks"
• Recovery needs: "what was I doing", "where was I", "got interrupted", "lost my train of thought"
• Transition language: "switching to", "coming back to", "need to pause this"
• Memory externalization: "help me remember", "keep track of", "don't let me forget"

Core Approach

1. Capture Current State Quickly

The faster and lower-friction the capture, the more likely it gets used. Minimize questions—get essentials now, details later.

2. Make Context Self-Explanatory

Future-you (or interrupted-you) shouldn't need to reconstruct meaning. Capture not just what, but enough why and where.

3. Track Open Loops

Human working memory fails with unfinished business. Explicitly note pending decisions, unanswered questions, things-to-remember.

4. Enable Clean Transitions

Good context handoff means you can fully engage with the new task without background worry about the old one.

Context Save Templates

Quick Save (30 seconds)

Minimum viable context capture:

## [Task Name] - Checkpoint [timestamp]

Current focus: [one sentence - what you're working on right now]
Next action: [the very next thing to do when resuming]
Open loops: [any pending decisions or questions]

Standard Save (2 minutes)

More comprehensive:

## [Task Name] - Context Save [timestamp]

### Current State
- Working on: [specific component/section/problem]
- Progress so far: [what's done]
- Current thought: [where your head was at]

### Next Steps
1. [immediate next action]
2. [what comes after that]

### Open Loops
- Decisions needed: [choices to make]
- Questions: [things you were wondering about]
- To remember: [important context for later]

### Files/Resources
- Currently open: [relevant files]
- References: [docs/links you had up]

Interrupt Recovery Save

When interrupted mid-flow:

## [Task] - INTERRUPTED [timestamp]

Before interruption, I was:
- Doing: [exact action]
- About to: [next micro-step]
- Thinking: [current train of thought]

Resume cue: [one sentence to trigger memory]
Open tabs/context: [what was on screen]

Context Restore Protocols

"Where Was I?" Recovery

When user returns but doesn't remember state:

1. Find the checkpoint - Locate most recent save
2. Read "Resume cue" first - Often triggers full context
3. Orient - "You were [situation], working on [specific thing]"
4. Immediate action - "Your next step was: [action]"
5. Context load - Share open loops and progress

Smooth Resumption Pattern

Welcome back to [task].

You were: [working on X, had just completed Y]
Your next step was: [specific action]

Still relevant:
- [open loop 1]
- [open loop 2]

Would you like to continue from there, or has the situation changed?

Working Memory Externalization

Open Loops Tracker

Maintain a running list of:

• Decisions pending: "Still need to choose between A and B"
• Questions unanswered: "Waiting to hear about X"
• Things to remember: "Don't forget constraint Y applies"
• Half-formed thoughts: "Was wondering if Z approach might work"

Update this continuously during work, not just at save points.

Multi-Task Context Map

When juggling multiple things:

Active Context Map:

[Task A] - Status: [active/paused/blocked]
- Last action: [what]
- Next action: [what]
- Blocking on: [if applicable]

[Task B] - Status: [active/paused/blocked]
- Last action: [what]
- Next action: [what]
- Dependencies: [if applicable]

Clean Transitions

Closing the Loop Before Switching

Before switching from [Current Task] to [New Task]:

1. Current task state: [complete/paused/blocked]
2. Next time you work on this: [start here]
3. Open questions to capture: [list]
4. Clean break?: [anything unresolved that might nag you?]

Good to switch? [confirm nothing will leak into next task]

Handoff Template

When passing work to future-self or others:

## Context Handoff: [Task]

### What Was Accomplished
- [completed item 1]
- [completed item 2]

### Current State
[where things stand now]

### What's Next
[what should happen next and why]

### Important Context
[non-obvious things someone needs to know]

### Gotchas/Warnings
[things that might trip someone up]

Interrupt Recovery Techniques

The "Before I Was Interrupted" Prompt

Immediately after interruption (even brief):

• "Before I was interrupted, I was [doing X]"
• "I was about to [next micro-action]"
• "I was thinking [thought process]"

Captures volatile working memory before it decays.

Reconstruct From Artifacts

When no checkpoint exists:

• Last file saved: [what was the last edit about?]
• Open windows/tabs: [what was in context?]
• Recent searches: [what were you looking for?]
• Terminal history: [what commands were running?]

Piece together context from digital breadcrumbs.

The 60-Second Refresh

Quick context reload protocol:

1. Skim last checkpoint (10s)
2. Check open loops (10s)
3. Identify next action (10s)
4. Load relevant file/resource (15s)
5. Start doing next action (15s)

Gets you back in motion fast.

File-Based Context Management

Suggested Storage Pattern

.context/
  current.md           # Active work context
  checkpoints/         # Timestamped saves
    2025-11-06-1430.md
    2025-11-06-1520.md
  open-loops.md        # Running list of pending items
  task-map.md          # Multi-task overview

Or use whatever system the user prefers—the pattern matters more than the location.

Anti-Patterns to Avoid

❌ Don't make context saves so detailed they're burdensome (friction kills usage) ❌ Don't skip the "why" (future-you won't remember the reasoning) ❌ Don't leave open loops implicit (they'll nag subconsciously) ❌ Don't assume memory is reliable (it's not, especially after interruptions)

Success Indicators

This skill is working when:

• User can resume work quickly after interruptions
• User reports less anxiety about task switching
• User actually uses the checkpoints (low friction)
• "What was I doing?" becomes less frequent

Relationship to Other Skills

• getting-started: Can help resume momentum after context restore
• sensing-limits: Consider capacity when planning resume timing
• May store context in files that user can reference later