Daily Patterns Pack

Turn your daily notes into a system that learns how you work.

A two-tool feedback loop: capture what you do, discover what you should automate.

Demo Video

Why

You use Claude Code. You get things done. But then:

• Tomorrow, you forget what you decided today
• You repeat the same manual workflows weekly
• You never notice which tasks drain your time
• You can’t see your own work patterns

Claude doesn’t remember between sessions. Your daily notes do—but only if you write to them, and only if something reads them back.

Install

# In Claude Code:
/plugin marketplace add aplaceforallmystuff/marketplace
/plugin install daily-patterns-pack@aplaceforallmystuff

Use cases

• Use it when you finish a Claude Code session and want a structured record of what you did, decided, and changed.
• Use it when you make an important decision mid-session and want the rationale captured before you forget it.
• Use it when you switch contexts and want to checkpoint the work so far.
• Use it after two or more weeks of logs to find which manual workflows you repeat and should automate.
• Use it when you want to see your own time and productivity patterns—deep work versus admin, day-of-week clustering.
• Use it when you need concrete, evidence-backed skill or agent build recommendations instead of guesses.

How it works

Two tools. One feedback loop.

/log-to-daily — Capture

At the end of each session, run /log-to-daily. Claude analyzes your conversation and appends a structured log to today’s daily note:

## Session Log - 2:45 PM

**Focus:** Local LLM integration for cost optimization

### Completed
- Overhauled inbox-processor with local_classify
- Created keep-warm launchd agent for Ollama
- Tested classification accuracy (~66%)

### Decisions Made
| Decision | Rationale |
|----------|-----------|
| Hybrid local/cloud pattern | Local does mechanical work, Claude does judgment |
| 4-minute keep-warm interval | Prevents cold start without wasting resources |

### Files Modified
- [[inbox-processor]] - Added local LLM integration

### Next Steps
- [ ] Monitor token savings over next week

What gets captured:

• Concrete outcomes (not just “worked on stuff”)
• Decisions with rationale
• Files created/modified (with wiki links)
• Follow-up tasks

@vault-analyst — Discover

After a few weeks of logs, run @vault-analyst. It reads your daily notes and surfaces:

• Recurring tasks you do manually every week
• Time patterns (when you do deep work vs admin)
• Friction points (what blocks you repeatedly)
• Automation candidates (specific skills/agents to build)

@vault-analyst is read-only. It observes, analyzes, and recommends; it never modifies your vault.

The compounding effect

Week 1: You log sessions. Nothing special yet.

Week 4: Vault-analyst spots your Friday financial review pattern. You build a skill.

Week 8: That skill appears in your logs. Vault-analyst notices you always run it after checking email. Suggests an agent that coordinates both.

Week 12: Your system now handles things you used to do manually. Your logs shift to higher-level work. Vault-analyst notices that pattern too.

The more you log, the more patterns emerge. The more patterns, the better your automation. The better your automation, the more valuable work you log.

What makes good logs

Specific outcomes:

• ✅ “Created inbox-processor with local_classify integration”
• ❌ “Worked on inbox stuff”

Decisions with reasoning:

• ✅ “Chose hybrid local/cloud pattern because local model accuracy was only 66%”
• ❌ “Made some decisions about the approach”

Linked files:

• ✅ “Modified [[inbox-processor]] and [[vault-analyst]]”
• ❌ “Changed some files”

The quality of your logs determines the quality of vault-analyst’s recommendations.

Example

After two or more weeks of logs, ask the analyst what to automate:

@vault-analyst what should I automate?

It returns a report citing the specific patterns behind each recommendation. Illustrative output:

## Vault Analysis Report

**Analyzed:** 30 daily notes (Jan 1 - Jan 30)

### Automation Recommendations

**1. weekly-financial-sync** — Priority: High
- **Why:** "Sync MoneyWiz" appears in 4 of 4 Friday notes
- **What it would do:** Auto-pull transactions, categorize, update Notion
- **Time saved:** ~30 min/week
- **Effort to build:** Medium

**2. newsletter-preflight** — Priority: High
- **Why:** 6 session logs mention "check slop", "verify links", "SEO meta"
- **What it would do:** Run all pre-publish checks in sequence
- **Time saved:** ~15 min/issue
- **Effort to build:** Low (compose existing skills)

Other invocations:

# Full analysis of the last 30 days
@vault-analyst analyze my vault

# Time and productivity patterns
@vault-analyst what are my productivity patterns?

Configuration

Requirements:

• Daily notes in Obsidian — Format: 00 Daily/YYYY/YYYYMMDD.md (configurable)
• Claude Code — These are Claude Code skills/agents
• Consistency — The system learns from repetition. Sporadic logging = weak patterns.

Edit the daily note path in log-to-daily/SKILL.md:

Path: `00 Daily/YYYY/YYYYMMDD.md`

Adjust to match your vault structure.

Files included

daily-patterns-pack/
├── agents/
│   └── vault-analyst.md
├── skills/
│   └── log-to-daily/
│       └── SKILL.md
├── install.sh
├── LICENSE
└── README.md

Philosophy

Most AI tools optimize for single sessions. This pack optimizes for months.

Your daily notes become training data—not for an AI model, but for your personal automation system. Every session logged is a data point. Every pattern detected is an automation waiting to be built.

You’re not just using Claude. You’re teaching your vault to work like you do.

License

MIT — see LICENSE.

---

Part of the Claude Code plugin ecosystem. Start small, compound over time.