Skip to main content

/work-next - Comprehensive Work Discovery

Analyzes multiple data sources to help you understand current project state and interactively decide what work to begin next.

System Prompt

EXECUTION DIRECTIVE: When the user invokes this command, you MUST execute all 8 discovery steps in sequence, then present an interactive summary with recommendations.

Step 1: Recent Context Activity

Query the context database for recent work:

/cxq --recent 200

Extract:

  • Last 10-20 unique activities/topics
  • Most recent decisions made
  • Open questions or blockers mentioned

Step 2: Active Project Plan

Find and analyze the active project plan:

Glob("**/PILOT*.md")
Glob("**/PROJECT-PLAN*.md")
Read the most recent plan file

Extract:

  • Current phase/sprint
  • Active tracks and their status
  • Blocking items
  • Next milestones

Step 3: Tasklist Status

Check the tasklist for pending items:

Glob("**/TASKLIST*.md")
Read and parse task status

Extract:

  • In-progress tasks
  • Blocked tasks
  • Next pending tasks by priority
  • Recently completed (for context)

Step 4: Recent Git History

Analyze recent commits:

git log --oneline -20
git log --since="24 hours ago" --oneline

Extract:

  • What was worked on recently
  • Which areas have momentum
  • Any revert or fix commits (indicating issues)

Step 5: Uncommitted Changes

Check working tree status:

git status --short
git diff --stat

Extract:

  • Staged changes (ready to commit)
  • Unstaged modifications (work in progress)
  • Untracked files (new work)

Step 6: Directory Change Analysis

Identify which areas have changes:

git status --short | cut -c4- | xargs -I {} dirname {} | sort -u

Extract:

  • Which submodules have changes
  • Which directories are active
  • Spread of changes (focused vs scattered)

Step 7: Recent Session Analysis

Check Claude Code session files:

ls -lt ~/.claude/projects/-Users-halcasteel-PROJECTS-coditect-rollout-master/*.jsonl | head -5

For the most recent session:

  • Extract last 50-100 messages
  • Identify topics discussed
  • Find any TODO items or follow-ups mentioned

Step 8: Checkpoint Review

Check for recent checkpoints:

Glob("**/CHECKPOINT*.md")
ls -lt context-storage/checkpoints/ | head -5

Extract:

  • Last checkpoint date and summary
  • Time since last checkpoint
  • Whether checkpoint is needed

Interactive Output Format

After gathering all data, present:

╔══════════════════════════════════════════════════════════════════╗
║ WORK DISCOVERY SUMMARY ║
╠══════════════════════════════════════════════════════════════════╣

📊 PROJECT STATUS
├─ Phase: [Current phase from PILOT plan]
├─ Sprint: [Current sprint if applicable]
└─ Overall: [X]% complete

📋 ACTIVE WORK
├─ In Progress: [N tasks]
│ • [Task 1 - brief description]
│ • [Task 2 - brief description]
├─ Blocked: [N tasks]
│ • [Blocker 1]
└─ Uncommitted: [N files in M directories]
• [dir1/] - N files modified
• [dir2/] - N files added

🕐 RECENT ACTIVITY (Last 24h)
├─ Commits: [N commits]
├─ Topics: [topic1, topic2, topic3]
└─ Last session: [time ago] - [brief topic]

🎯 RECOMMENDED NEXT ACTIONS

[1] [HIGH] Continue: [in-progress task with momentum]
└─ Context: [why this is recommended]

[2] [HIGH] Commit: [uncommitted work description]
└─ Files: [N files ready to commit]

[3] [MEDIUM] Unblock: [blocked task]
└─ Blocker: [what's blocking]

[4] [MEDIUM] Start: [next pending task]
└─ From: [PILOT plan track]

[5] [LOW] Checkpoint: Create progress checkpoint
└─ Last: [time since last checkpoint]

╚══════════════════════════════════════════════════════════════════╝

Then ask:

Which would you like to work on? [1-5, or describe different task]

Usage

/work-next              # Full discovery with recommendations
/work-next --quick # Abbreviated version (steps 4-6 only)
/work-next --git # Git status focus only
/work-next --context # Context/session focus only

Examples

# Start of session - full discovery
/work-next

# Quick check during work
/work-next --quick

# After a break - what was I doing?
/work-next --context

Options

OptionDescription
--quickAbbreviated discovery (git + uncommitted only)
--gitFocus on git status and history
--contextFocus on context DB and sessions
--verboseShow raw data from each step
--no-recommendShow status without recommendations

Data Sources

SourceLocationPurpose
Context DBcontext-storage/sessions.db + org.db (ADR-118)Recent activities, decisions
Project Plan**/PILOT*.md, **/PROJECT-PLAN*.mdCurrent phase, tracks
Tasklist**/TASKLIST*.mdTask status
Git Log.git/Recent commits
Git StatusWorking treeUncommitted changes
Sessions~/.claude/projects/*.jsonlSession history
Checkpointscontext-storage/checkpoints/Progress snapshots

Integration with Other Commands

After /work-nextUse
Selected task/agent <name> "task" to execute
Need to commitgit add . && git commit
Need checkpoint/cx to capture context
Need more context/cxq "specific query"
Start fresh session/orient

# 1. Start of day/session
/work-next

# 2. Select recommended action or specify your own
> 1 # Continue in-progress work

# 3. Work on the task...

# 4. Before break/end
/work-next --quick # Check if anything to commit

# 5. End of session
/cx # Capture context

Success Output

When discovery completes successfully:

✅ COMMAND COMPLETE: /work-next
Data sources: 8/8 analyzed
Recommendations: N prioritized actions
Interactive: Ready for selection

Completion Checklist

Before marking complete:

  • All 8 discovery steps executed
  • Summary displayed with status
  • Recommendations prioritized
  • Interactive prompt shown

Failure Indicators

This command has FAILED if:

  • ❌ Context database unavailable
  • ❌ No project plan found
  • ❌ Git repository not accessible
  • ❌ Session files not readable

When NOT to Use

Do NOT use when:

  • Already know what to work on
  • Quick single task (overhead not worth it)
  • Not in a CODITECT project

Anti-Patterns (Avoid)

Anti-PatternProblemSolution
Skip recommendationsMiss prioritiesReview all options
Ignore blockersStuck laterAddress blockers first
Run repeatedlyWasted timeRun once at session start

Principles

This command embodies:

  • #9 Based on Facts - Uses multiple data sources
  • #10 Research When in Doubt - Comprehensive discovery
  • #3 Complete Execution - All 8 steps executed

Full Standard: CODITECT-STANDARD-AUTOMATION.md


Version: 1.0.0 Created: 2026-01-03 Author: CODITECT Team