Lot

das Lot — the plumb line. Finds true vertical.

A Claude Code plugin that helps you track what matters, notice when you drift, and stay aligned with your goals. Works for individuals and small teams (2-10 people).

Lot is a personal operating system built on Markdown files. It composes 40+ existing Claude Code skills into five core workflows:

• Onboard — scans your environment, interviews you about what matters, builds a profile progressively. Quit anytime — every answer creates something usable.
• Check-in — 2-minute morning pulse: energy, intention, blockers, habits. Calibrated to your profile and tone preference.
• Inbox — collects incoming signals from email, Linear, and daily notes. Interactive triage. Self-improving eval mode that proposes system improvements after 20+ triaged items.
• Goals — OKR and behavioral goal tracking with lightweight drift detection. Compares declared priorities against actual activity.
• Config — tone (direct / coaching / neurodivergent), depth, framework, channel, rhythm. Changeable anytime.

# Add as a local marketplace
claude plugin marketplace add ~/ai_projects/lot

# Install
claude plugin install lot@lot-marketplace

Start a new Claude Code session and run:

/lot

Lot stores everything as Markdown with flat YAML frontmatter in lot-data/ inside your working directory. If you use Obsidian, that means Dataview queries work out of the box.

lot-data/
├── config.md              # Your settings
├── profile.md             # Your profile (filled progressively)
├── habits-registry.md     # Trackable habits + activation energy levels
├── scan.md                # Latest environment scan
├── inbox/                 # Incoming signals
│   ├── processed/         # Triaged items
│   └── eval/              # Self-improvement reports
├── sessions/              # Onboard + triage transcripts
├── shared/
│   ├── goals/             # OKR and behavioral goals
│   └── checkins/          # Daily check-ins
└── private/               # Never synced

Every file has an id matching its filename, a type field, and a visibility field (shared/private). Team members share shared/ via git; private/ stays local.

Three settings shape how Lot interacts with you:

Tone — how questions are phrased:

• direct — short, no framing
• coaching — warmer, reflective
• neurodivergent — shortest possible, explicit stop checkpoints

Depth — how much onboarding collects:

• quick (~2 min) — one goal, basic profile
• standard (~5 min) — goal + rhythm + friction
• deep (~15 min) — full profile with episodes and tensions

Channel — where check-ins happen:

• terminal — only when you run /lot
• telegram — proactive pings at configured times
• both — telegram for pings, terminal for depth

Lot tracks habits with activation energy levels — on low-energy days, it suggests Level 1 alternatives (cold water on face, 5-min walk) instead of the full habit. Habits link to goals, so drift detection knows when you're missing the behaviors that support your priorities.

Lot skills work across multiple runtimes. AGENTS.md contains all documentation in agent-agnostic format. CLAUDE.md points to it.

1. Welcome first, scan later — explain what's happening before running anything
2. Every answer writes a file — no buffered state, interrupt-safe
3. Coverage, not script — stop at any depth, get proportional value
4. Show where data comes from — every scan finding labels its source
5. AskUserQuestion for closed lists — structured choices, not free text guessing
6. Concrete over abstract — one follow-up for vague answers, then accept
7. Facts over judgment — drift detection states facts, user interprets

MIT

Gleb Kalinin — github.com/glebis