work-report

name	work-report
description	Generate human-readable work reports for standups, team syncs, and manager updates. Summarizes accomplishments, progress, and next steps across projects.
arguments

/work-report

Generate work reports for human communication - standups, team meetings, manager syncs.

When to Use

• Daily standup preparation
• Weekly team sync
• Progress update for manager/PM/TPM
• Reflecting on what was accomplished
• Preparing status for cross-functional meetings

Usage

/work-report                    # Smart default (today or recent)
/work-report today              # Today's work
/work-report yesterday          # Yesterday's work
/work-report this week          # This week
/work-report last week          # Last week
/work-report Jan 27-31          # Custom range

Output Philosophy

Human-first: Written for people, not systems. Focus on what was accomplished in plain language.

Audience-agnostic: Start neutral, user can request adjustments:

• "Make it more technical for engineering sync"
• "Summarize for PM - focus on deliverables"
• "Add more detail on the auth work"

Divide & conquer: High-level summary first, details only for significant milestones/deliverables/findings.

Implementation

Step 1: Determine Time Period

Parse the user's request to determine the time range:

• "today" → today's date
• "yesterday" → yesterday's date
• "this week" → Monday to today
• "last week" → Previous Monday to Sunday
• Custom range → Parse dates

Default (no argument): Today if there's activity, otherwise last 7 days.

Step 2: Gather Project Data

# Read project index
cat ~/.varie/manager/projects.yaml

Step 3: Gather Session Data

# Find all usage files
ls ~/.varie/sessions/*/usage.json

# Read each and filter by time period
# Group by initial_cwd to identify which projects had activity

For each usage.json, check if started_at or last_updated falls within the time period.

Step 4: Detect Meaningful Work

Why: Token count is a poor proxy for meaningful work. A 500-token session with a bug fix commit is more valuable than a 50K-token exploration with no output. Filter noise to report on actual accomplishments.

Work Categories

Category	Signal	Report As
COMPLETED	Has git commits in time period	"Done" section
IN_PROGRESS	Has uncommitted code changes (excluding noise)	"In Progress" section
RESEARCH	High tokens (>10K) + handover/checkpoint exists	"Exploring" section
EXCLUDED	Noise (see below)	Not reported (footnote only)

Noise Detection (EXCLUDED)

Exclude from report:

• CLAUDE.md Varie-section only: Auto-injected plugin reference, not real work
• Minimal sessions: <1K tokens with no file changes
• Quick opens: started_at ≈ last_updated (within 60 seconds)

Git-Based Detection

For each project with sessions in time period:

# 1. Check for commits in period
git log --oneline --since="$START_DATE" --until="$END_DATE"

# 2. Check for uncommitted changes
git diff --name-only HEAD
git status --porcelain | grep '^\?\?'  # untracked files

# 3. Detect CLAUDE.md Varie-section-only change
# The auto-injected section is exactly:
#   ## Workstation
#   (empty line)
#   Key skills: `/work-start`, `/work-checkpoint`, `/work-handover`, `/work-status`.
# If CLAUDE.md diff contains ONLY this (≤4 lines added at end), it's noise.

Multi-Repo Session Handling

When initial_cwd is a parent directory (e.g., /workplace/projects/):

1. Primary: Check git activity across all known projects during session timeframe
   • Attribute session to projects that had commits or uncommitted changes
   • Same session can credit multiple projects
2. Fallback: If no git changes detected, mark as "Cross-project work"
   • Report under separate "Cross-Project" section
   • Suggest using /work-start <repo> <task> for better tracking

Step 5: Gather Work Details

For each project classified as COMPLETED, IN_PROGRESS, or RESEARCH:

# Git commit messages (for COMPLETED)
git log --oneline --since="$START_DATE" --until="$END_DATE"

# Find latest archive entry (for context)
LATEST=$(ls -t <repo>/archive/ 2>/dev/null | head -1)

# Read HANDOVER.md for recent work
cat <repo>/archive/$LATEST/HANDOVER.md 2>/dev/null

# Check for issues
cat <repo>/issues/INDEX.md 2>/dev/null

Step 6: Synthesize Report

Transform technical data into human-readable prose. Write as if speaking in standup - conversational, focused on what matters.

## Work Report - [Time Period]

### Summary
[1-2 sentence high-level summary - what was the focus, what got done]

### Done
- **[Project]**: [Descriptive accomplishment - what was done, why it matters]
- **[Project]**: [Another accomplishment]

### In Progress
- **[Project]**: [Current focus] - [brief status, e.g., "working through API integration"]

### Exploring
- **[Project]**: [What was investigated and key takeaways]

### Cross-Project
- [Only if multi-repo sessions with no clear attribution]
- [Suggest `/work-start <repo> <task>` for better tracking]

### Key Findings / Decisions
- [Important discoveries, blockers resolved, architectural decisions]
- [Only include if significant - skip section if nothing notable]

### Next Steps
- [What's planned next - actionable items]

---
*[N] sessions across [M] projects ([X] excluded as noise)*

Section rules:

• Omit empty sections entirely (don't show "Exploring" if none)
• "Done" requires git commits - don't inflate with uncommitted work
• "Cross-Project" only appears when attribution unclear
• Footer notes excluded sessions so user knows filtering happened

Example Output

Daily Report

## Work Report - February 3, 2026

### Summary
Shipped several bug fixes for Workstation dispatch and made progress on character extension i18n.

### Done
- **varie-workstation**: Fixed the dispatch Enter key bug that was causing commands to fail intermittently
- **varie-workstation**: Added external repo support to project discovery

### In Progress
- **react-app**: Updating internationalization - working through translation key updates

### Exploring
- **data-pipeline**: Getting familiar with the architecture for potential integration work

### Next Steps
- Finish i18n updates and test across languages
- Document data-pipeline findings if integration moves forward

---
*6 sessions across 3 projects (2 excluded as noise)*

Weekly Report

## Work Report - Week of January 27, 2026

### Summary
Focused week on Workstation plugin development. Completed the visibility layer (Phase C2) and context tracking features (Phase C3). Set foundation for reporting.

### Done
- **varie-workstation**: Phase C2 - Project visibility via `/projects`, `/project` skills
- **varie-workstation**: Phase C3 - Context tracking, checkpoint hints, `/work-stats`
- **varie-workstation**: Auto-setup flow for status line configuration
- **react-app**: Fixed component loading race condition

### In Progress
- **varie-workstation**: Phase C4 - Reporting features

### Exploring
- **ml-pipeline**: Investigated optimization options - documented in handover

### Key Findings
- Repo-centric data model (archive/ in repos) better than session-centric for persistence
- Manager should coordinate, not command - sessions are autonomous

### Issues Resolved
- #003: Orchestrator terminal not opening

### Issues Opened
- #007: Auto-download workstation app on plugin install

### Next Steps
- Complete reporting skill
- Begin workstation app distribution setup
- Resume react-app API work

---
*12 sessions across 4 projects (5 excluded as noise)*

Handling Edge Cases

No Activity in Time Period

## Work Report - [Time Period]

No tracked sessions found for this time period.

To track work:
1. Ensure the varie-workstation plugin is installed
2. Status line should be configured (check with `/work-stats`)
3. Start working - sessions are tracked automatically

All Sessions Excluded (Only Noise)

## Work Report - [Time Period]

No meaningful work detected for this time period.

8 sessions were excluded:
- 6 had no code changes (CLAUDE.md updates only)
- 2 were brief opens (<1K tokens)

If you did meaningful work, consider:
- Using `/work-checkpoint` to save progress
- Using `/work-handover` to document exploration

Research Without Handover

If high-token session (>10K) with no commits, no uncommitted changes, and no handover:

### Exploring
- **[project]**: Exploration session (undocumented) - consider `/work-handover` to capture findings

No Handover Data for Completed Work

If a project has commits but no HANDOVER.md:

### Done
- **[project]**: [commit message summary from git log]

Git commits provide enough context; handover is optional for completed work.

Multi-Repo Session Without Clear Attribution

### Cross-Project
- **General exploration** (from /workplace/projects/): 2 sessions, 45K tokens
  - Tip: Use `/work-start <repo> <task>` for better tracking

Follow-up Refinements

After generating a report, user can request adjustments:

• "Make it shorter - just bullet points"
• "Add more detail on the workstation work"
• "Format for Slack"
• "Make it suitable for my manager (non-technical)"
• "Focus on blockers and risks"

Notes

• Reports are generated, not stored - run again to get updated data
• Meaningful work detection: Reports filter out noise (CLAUDE.md-only changes, minimal sessions) to focus on actual accomplishments
• Token/cost stats available via /work-stats if user wants technical details
• For detailed project status, use /project <name>
• For live session list, use /work-sessions
• Git commits are the primary signal for "Done" - uncommitted work shows as "In Progress"