Type /tot at any point. Claude branches into a few options, builds and scores them, and keeps the winner.
| Version | What it is | Open |
|---|---|---|
| v1.7 | Dark neon graphic style: one poster, then steps and example | Open v1.7 |
| v1.6 (this page) | Knowledge-sheet style: concepts, arguments, four diagrams | You are here |
| v1.5 | Original long page with the option picker and the picker logic | Open v1.5 |
| v1.3 | Short page: title, four steps, one example | Open v1.3 |
/tot is a planned global Claude Code skill based on Tree of Thoughts. When you type it, Claude reads the current task and suggests 3 or 4 different approaches, each with pros, cons, cost and a scoring method. You pick, Claude builds and scores each one, drops weak branches, backs up when one fails, and keeps the best. It matters because choosing between valid approaches by gut feeling is slow and easy to regret, and this makes the choice measurable.
One distinct approach to the task, not a small tweak of another.
A number from a test or measurement that rates a branch.
Drop a branch that scores too low or cannot be measured.
Return to an earlier branch when the current one fails.
The step where Claude shows 3 or 4 options and you choose.
The best-scoring branch, kept and applied as the default.
~/.claude/skills/.From your task to the kept winner
flowchart LR
subgraph Input
A["Your task and project files"]:::data
end
subgraph Process
B["Suggest 3 or 4 options"]:::process --> C["You pick"]:::process --> D["Build and score each"]:::process
end
subgraph Output
E["Winner kept as default"]:::solution
end
A --> B
D --> E
D -.->|"failed: backtrack"| C
classDef data fill:#e1bee7,stroke:#7b1fa2,color:#1a1a1a
classDef process fill:#fff3e0,stroke:#f57c00,color:#1a1a1a
classDef solution fill:#c8e6c9,stroke:#2e7d32,color:#1a1a1a
What a run is made of
classDiagram
class Run {
+String mode
+String scoringRule
}
class Option {
+String name
+String pros
+String cons
+String cost
+Number score
}
class Winner {
+String appliedAs
}
Run "1" --> "3..4" Option : builds
Option --> Winner : best score becomes
The settings and choices in the skill
mindmap
root(("/tot"))
Options per run
Suggest 3
Maximum 4
Run mode
One after another
Parallel sessions
Scoring by project
Audio clipping and loudness
Web app tests and screenshots
Bug hunt reproduce then fix
Design side by side
Setup
One global skills folder
The problem, the skill that answers it, and what it produces
graph TB
P["Many valid approaches, picked by gut"]:::problem
S["/tot skill"]:::solution
C1["Branch"]:::concept
C2["Score"]:::concept
C3["Backtrack"]:::concept
PR["Build and measure each option"]:::process
D["Scores and winner"]:::data
P -->|addressed by| S
S -->|uses| C1
S -->|uses| C2
S -->|uses| C3
C1 -->|feeds| PR
C2 -->|measured in| PR
PR -->|produces| D
C3 -->|triggered by failure in| PR
classDef problem fill:#ffcdd2,stroke:#c62828,color:#1a1a1a
classDef solution fill:#c8e6c9,stroke:#2e7d32,color:#1a1a1a
classDef concept fill:#e3f2fd,stroke:#1976d2,color:#1a1a1a
classDef process fill:#fff3e0,stroke:#f57c00,color:#1a1a1a
classDef data fill:#e1bee7,stroke:#7b1fa2,color:#1a1a1a
Reads your task and files, then lists every approach it can think of.
Merges near-duplicates so each option is a different idea.
Checks fit to the goal, measurability, cost and risk. Drops the weakest.
Top 3, up to 4, with pros, cons, cost and the scoring method.
Counts are examples. An option Claude cannot measure is always dropped.
| Option | Result | Score of 10 |
|---|---|---|
| Meow lead over techno, echo 28 | Kept as default | 9.2 |
| Meow plus Bhangra beat | Runner-up, backtrack target | 7.1 |
| Meow lead over techno, echo 18 | Melody a bit dry | 7.9 |
| Meow lead over techno, echo 40 | Clips, dropped | n/a |
| Ambient, heavy reverb | Melody buried, dropped | n/a |
Illustration only, not a real kraut-deck run.