Planned Claude Code skill · Tree of Thoughts on demand

/tot: from many options to one measured winner

Type /tot while you work. Claude reads your project, suggests a few approaches, you pick, and it builds, scores and keeps the best one.

YOUR WORK YOUR TASKwhat you are doing now PROJECT FILESopen files and code PROJECT TYPEaudio, web app, bug hunt RECENT HISTORYcommits and notes CLAUDE SKILL: /TOT branch · score · keep SUGGEST 3–4 BUILD + SCORE STEP 2 · YOU CHOOSE OPTION PICKER tick one, several or all A · MEOW LEAD + TECHNOcost low B · MEOW + BHANGRA BEATcost medium C · AMBIENT, HEAVY REVERBcost low D · CHIPTUNE MEOWcost high STEP 3 · CLAUDE BUILDS AND SCORES /TOT MEOW + TECHNO 8.6 MEOW + BHANGRA 7.1 AMBIENT pruned CHIPTUNE 6.4 echo 18 · 7.9 28 · 9.2 echo 40 · clips DEFAULT winner kept
Your work
  • Your tasknow
  • Project filescode
  • Project typeaudio / web / bug
  • Recent historycommits
↓

Claude skill: /tot

branch · score · keep
↓ suggest 3–4
Step 2 · you choose
  • ☑ A · Meow lead + technolow
  • ☑ B · Meow + Bhangramedium
  • ☐ C · Ambient reverblow
  • ☐ D · Chiptune meowhigh
↓ build + score
Step 3 · Claude builds and scores
  • A · echo 289.2 kept
  • A · echo 187.9
  • B · Bhangra7.1
  • A · echo 40clips
The flow

Five steps, one command

  1. Claude

    Read context

    Task, open files, project type, recent commits.

  2. Claude

    Suggest 3–4

    Each with pros, cons, cost and how it will be scored.

  3. You

    Choose

    Tick one, several, or all.

  4. Claude

    Build + score

    One after another, or in parallel sessions.

  5. Claude

    Keep winner

    Apply the best. Backtrack if it fails.

Inside step 2

How the agent picks options

  1. 8
    Brainstorm

    Lists every approach it can think of.

  2. 5
    Make them differ

    Merges near-duplicates.

  3. 4
    Rate each

    Runs the four checks below.

  4. 3
    Show you

    Top 3, up to 4.

fits the goalcan be measuredcostrisk

Counts are examples. An option it cannot measure is always dropped.

Inside step 3 · kraut-deck example

Scored arrangements

OptionResultScore / 10
Meow + techno, echo 28Kept as default9.2
Meow + techno, echo 18Melody a bit dry7.9
Meow + Bhangra beatBacktrack target7.1
Meow + techno, echo 40Clips, droppedn/a
Ambient, heavy reverbMelody buriedn/a

Illustration only, not a real run.

Auto-adapts

Scoring changes per project

AudioPeak level, clipping, loudness, CPU.
Web appTests pass, Lighthouse, screenshots.
Bug huntReproduce it, then prove the fix.
DesignSide-by-side preview, you judge.
Before we build

Open choices

  • Options per runsuggest: 3, max 4
  • Default run modesuggest: one after another
  • Who writes the scoring rulessuggest: Claude, you edit
  • Commit the winner automaticallysuggest: ask first
  • Setupone folder: ~/.claude/skills/tot/