openclaw-executive-assistant-local

# OpenClaw Executive Assistant (Local) > Skill by [ara.so](https://ara.so) — Hermes Skills collection. ## Overview This project provides a local-first OpenClaw workflow system for building AI-powered executive assistant capabilities. It focuses on three core operations: 1. **Data intake review** - Transform unknown files into structured intake reports 2. **Operational memory** - Convert work residue into daily logs and weekly summaries 3. **Offline communications triage** - Process exported emails into action lists All operations use local files only, produce reviewable markdown artifacts, and require no live integrations. ## Repository Structure ``` code-along/ ├── INDEX.md ├── 01-data-intake-review/ │ ├── incoming/ # Files to inspect │ ├── prompts/ # Prompt templates │ ├── outputs/ # Generated reports │ └── expected/ # Reference outputs ├── 02-operational-memory/ │ ├── inbox/ # Notes and work residue │ ├── prompts/ # Daily/weekly prompts │ ├── outputs/ # Generated logs │ └── schedule/ # Cron examples ├── 03-offline-communications-triage/ │ ├── eml/ # Exported email files │ ├── prompts/ # Triage prompts │ ├── outputs/ # Triage reports │ └── expected/ # Reference outputs └── mission-control/ # Dashboard (optional) ``` ## Exercise 1: Data Intake Review ### Purpose Transform unknown files in an `incoming/` folder into a trustworthy intake report. ### Workflow 1. **Place files to review:** ```bash # Add files to the incoming folder cp ~/Downloads/unknown-file.pdf code-along/01-data-intake-review/incoming/ ``` 2. **Use the intake review prompt:** The prompt template is in `prompts/intake-review.md`. Pass it to your AI assistant along with the contents of `incoming/`: ```markdown # Intake Review Prompt You are an executive assistant performing a data intake review. Review all files in the incoming/ folder and produce a report that includes: 1. File inventory (name, type, size, date) 2. Content summary for each file 3. Suggested categorization 4. Action items or next steps 5. Priority flags (urgent, routine, archive) Output format: Markdown Output location: outputs/intake-review.md ``` 3. **Expected output structure:** ```markdown # Data Intake Review *Generated: YYYY-MM-DD* ## Summary - Total files: X - Urgent items: Y - Requires action: Z ## File Inventory ### [filename.ext] - **Type:** Document/Image/Data - **Size:** XXX KB - **Date:** YYYY-MM-DD - **Summary:** Brief content description - **Category:** Work/Personal/Archive - **Action:** Review/File/Respond - **Priority:** High/Medium/Low [Repeat for each file] ## Recommended Actions 1. ... 2. ... ``` ## Exercise 2: Operational Memory ### Purpose Convert daily work residue into structured logs and weekly summaries. ### Daily Log Workflow 1. **Add work residue to inbox:** ```bash # Add notes, snippets, or quick captures echo "Met with design team - new mockups ready" > code-along/02-operational-memory/inbox/note-$(date +%Y%m%d).txt ``` 2. **Use the daily log prompt** (`prompts/daily-log.md`): ```markdown # Daily Log Prompt You are an executive assistant creating a daily work log. Review all items in the inbox/ folder from today and produce: 1. **Date header** 2. **Wins** - Completed items 3. **Progress** - Items in motion 4. **Blockers** - Issues or delays 5. **Tomorrow** - Planned next actions 6. **Notes** - Observations or reminders Output format: Markdown Output location: outputs/daily-log.md Filename pattern: daily-YYYY-MM-DD.md ``` 3. **Expected daily log output:** ```markdown # Daily Log: 2026-05-11 ## Wins - ✅ Completed data intake review system - ✅ Shipped v2.1 of client dashboard ## Progress - 🔄 OpenClaw workshop prep (80% complete) - 🔄 Q2 planning document (draft stage) ## Blockers - ⚠️ Waiting on legal review for contract - ⚠️ Need API keys from DevOps ## Tomorrow - [ ] Finalize workshop slides - [ ] Review Q2 budget proposal - [ ] Team sync at 2pm ## Notes - Design team shared new mockups in Figma - Consider async standup format for remote team ``` ### Weekly Summary Workflow **Use the weekly hype prompt** (`prompts/weekly-hype.md`): ```markdown # Weekly Hype Prompt You are an executive assistant creating a weekly summary. Review all daily logs from this week (outputs/daily-*.md) and produce: 1. **Week of [date range]** 2. **Highlights** - Major wins and milestones 3. **Momentum** - Projects advancing 4. **Attention needed** - Recurring blockers 5. **Next week focus** - Priorities for the week ahead 6. **Metrics** (optional) - Quantifiable progress Output format: Markdown Output location: outputs/weekly-hype.md Filename pattern: weekly-YYYY-Www.md ``` ### Automation Example Schedule daily log generation with cron (see `schedule/cron-examples.md`): ```bash # Run daily at 6pm 0 18 * * * cd ~/openclaw-assistant && ./generate-daily-log.sh # generate-daily-log.sh example: #!/bin/bash DATE=$(date +%Y-%m-%d) AI_PROMPT=$(cat code-along/02-operational-memory/prompts/daily-log.md) # Pass inbox contents and prompt to your AI CLI tool # ai-cli "$AI_PROMPT" --context "code-along/02-operational-memory/inbox/*" \ # > "code-along/02-operational-memory/outputs/daily-$DATE.md" ``` ## Exercise 3: Offline Communications Triage ### Purpose Process exported email files into structured action lists. ### Workflow 1. **Export emails to .eml format:** ```bash # Place exported emails in the eml/ folder cp ~/exported-emails/*.eml code-along/03-offline-communications-triage/eml/ ``` 2. **Use the email triage prompt** (`prompts/email-triage.md`): ```markdown # Email Triage Prompt You are an executive assistant performing email triage. Review all .eml files in the eml/ folder and produce: 1. **Urgent** - Requires immediate response 2. **Action Required** - Needs response (24-48h) 3. **FYI** - Informational, no action needed 4. **Delegate** - Should be handled by someone else 5. **Archive** - Safe to file away For each email include: - From/Subject - Brief summary - Suggested response or action - Priority level Output format: Markdown Output location: outputs/email-triage.md ``` 3. **Expected triage output:** ```markdown # Email Triage Report *Generated: YYYY-MM-DD HH:MM* ## Urgent (Response Today) ### From: client@example.com | Re: Production Issue - **Summary:** Database timeout errors affecting users - **Action:** Coordinate with DevOps for immediate fix - **Priority:** 🔴 Critical ## Action Required (24-48h) ### From: legal@company.com | Re: Contract Review - **Summary:** Q2 vendor contract needs signature - **Action:** Review terms, sign if acceptable - **Priority:** 🟡 High ## FYI (No Action) ### From: team@company.com | Re: Weekly Newsletter - **Summary:** Company updates and team wins - **Action:** None - informational - **Priority:** 🟢 Low ## Delegate ### From: recruiter@agency.com | Re: Candidate Pipeline - **Summary:** Three candidates ready for interviews - **Action:** Forward to hiring manager Sarah - **Priority:** 🟡 Medium ## Archive [Emails that can be filed with no action] ``` ## Common Patterns ### Pattern 1: Copy-Paste Workflow ```markdown 1. Open AI assistant (Claude, ChatGPT, etc.) 2. Copy prompt from prompts/*.md 3. Attach or paste relevant files from incoming/inbox/eml/ 4. Run generation 5. Save output to outputs/*.md 6. Review and edit as needed ``` ### Pattern 2: Scripted Automation ```bash #!/bin/bash # automated-intake.sh PROMPT=$(cat code-along/01-data-intake-review/prompts/intake-review.md) FILES=$(ls code-along/01-data-intake-review/incoming/*) # Use your AI CLI tool of choice # ai-cli "$PROMPT" --files "$FILES" > outputs/intake-review-$(date +%Y%m%d).md ``` ### Pattern 3: Scheduled Heartbeat ```bash # Add to crontab # Daily log at 6pm weekdays 0 18 * * 1-5 cd ~/openclaw-assistant && ./daily-log.sh # Weekly summary Friday at 5pm 0 17 * * 5 cd ~/openclaw-assistant && ./weekly-summary.sh ``` ## Configuration ### Environment Setup Create a `.env` file for AI API configuration: ```bash # .env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} OPENAI_API_KEY=${OPENAI_API_KEY} AI_MODEL=claude-3-5-sonnet-20241022 ``` ### Prompt Customization Edit prompt files to match your workflow: ```bash # Customize intake review categories nano code-along/01-data-intake-review/prompts/intake-review.md # Adjust daily log sections nano code-along/02-operational-memory/prompts/daily-log.md # Modify email triage buckets nano code-along/03-offline-communications-triage/prompts/email-triage.md ``` ## Troubleshooting ### Issue: AI output not matching expected format **Solution:** Add explicit format instructions to prompts: ```markdown CRITICAL: Output must be valid Markdown with exactly these sections: - Summary - File Inventory - Recommended Actions Use ## for section headers. Use - for bullet lists. ``` ### Issue: Large files causing context limits **Solution:** Process in batches: ```bash # Split incoming files into chunks for file in incoming/*.pdf; do # Process individually echo "Processing $file..." done ``` ### Issue: Automation script not running **Solution:** Check cron logs and permissions: ```bash # View cron logs grep CRON /var/log/syslog # Ensure scripts are executable chmod +x *.sh # Test script manually ./daily-log.sh ``` ### Issue: Email .eml parsing errors **Solution:** Ensure proper export format from email client. Most clients support "Save as .eml" or "Export to file" options. If parsing fails, extract plain text first: ```bash # Extract text from .eml grep -A 1000 "^$" email.eml | tail -n +2 > email.txt ``` ## Integration Tips ### With Obsidian/Notion ```bash # Symlink outputs to your notes folder ln -s ~/openclaw-assistant/code-along/02-operational-memory/outputs ~/Obsidian/Daily-Logs ``` ### With Git for Versioning ```bash # Track generated logs cd code-along/02-operational-memory/outputs git init git add daily-*.md weekly-*.md git commit -m "Daily log archive" ``` ### With Markdown Viewers ```bash # Serve outputs as local site cd code-along python -m http.server 8000 # Open http://localhost:8000 ``` ## Best Practices 1. **Review before archiving** - Always human-review AI outputs before filing 2. **Consistent naming** - Use ISO date formats (YYYY-MM-DD) in filenames 3. **Regular cleanup** - Archive old logs monthly to keep folders manageable 4. **Prompt iteration** - Refine prompts based on output quality over time 5. **Local-first** - Keep sensitive data local; only upload sanitized examples