Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Worktree-first development environment — create per-issue or per-PR git worktrees, optionally with tmux sessions, so multiple work streams don't collide.
.claude/skills/evolution-foundation-dev-project-session-manager/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 778% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 55% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 621% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -32% | 0% |
Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.
Worktree-first development. Create per-issue or per-PR git worktrees so multiple work streams (different bugs, different features) don't collide on the same checkout.
bash# Per issue git worktree add ../{repo}-issue-123 -b issue/123-fix-auth # Per PR (review) git worktree add ../{repo}-pr-456 origin/feature/foo # Per experiment git worktree add ../{repo}-experiment-perf -b experiment/perf
bashtmux new-session -d -s {repo}-issue-123 -c ../{repo}-issue-123
bashgit worktree list
bashgit worktree remove ../{repo}-issue-123 git branch -d issue/123-fix-auth # if merged
{repo}-issue-{N} — issue work{repo}-pr-{N} — PR review{repo}-experiment-{name} — experiments{repo}-hotfix-{name} — urgent fixesSave the active worktree map to workspace/development/research/[C]worktrees-{date}.md:
markdown## Active Worktrees — {date} | Path | Branch | Purpose | tmux session | |---|---|---|---| | ../evo-ai-issue-123 | issue/123 | fix auth bug | evo-ai-issue-123 | | ../evo-ai-pr-456 | origin/feat/x | review | (none) |
@flow-git (when worktrees need rebasing or commit splitting)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 5,423 | 11,104 | +105% | 1 | 1 | 0% | 227 | 1,993 | +778% | 0 | 0 | — |
case-02 | fail→pass | 7,028 | 9,735 | +39% | 1 | 1 | 0% | 1,260 | 1,958 | +55% | 0 | 0 | — |
case-03 | fail→pass | 4,525 | 5,981 | +32% | 1 | 1 | 0% | 246 | 1,773 | +621% | 0 | 0 | — |
case-04 | pass→pass | 6,299 | 2,348 | -63% | 1 | 1 | 0% | 1,070 | 975 | -9% | 0 | 0 | — |
case-05 | pass→pass | 11,790 | 6,194 | -47% | 1 | 1 | 0% | 1,986 | 1,522 | -23% | 0 | 0 | — |
case-06 | pass→pass | 6,416 | 4,440 | -31% | 1 | 1 | 0% | 1,167 | 1,331 | +14% | 0 | 0 | — |
case-07 | fail→pass | 5,809 | 4,114 | -29% | 1 | 1 | 0% | 1,006 | 1,332 | +32% | 0 | 0 | — |
case-08 | fail→pass | 9,413 | 3,367 | -64% | 1 | 1 | 0% | 1,711 | 1,172 | -32% | 0 | 0 | — |
case-09 | fail→pass | 7,751 | 3,450 | -55% | 1 | 1 | 0% | 1,297 | 1,156 | -11% | 0 | 0 | — |
case-10 | fail→fail | 9,227 | 4,657 | -50% | 1 | 1 | 0% | 1,524 | 1,430 | -6% | 0 | 0 | — |
case-11 | pass→pass | 5,868 | 2,138 | -64% | 1 | 1 | 0% | 900 | 965 | +7% | 0 | 0 | — |
case-12 | pass→pass | 9,407 | 6,627 | -30% | 1 | 1 | 0% | 1,708 | 1,625 | -5% | 0 | 0 | — |
case-13 | fail→pass | 8,451 | 4,943 | -42% | 1 | 1 | 0% | 1,519 | 1,391 | -8% | 0 | 0 | — |
case-19 | pass→pass | 5,791 | 3,836 | -34% | 1 | 1 | 0% | 1,019 | 1,207 | +18% | 0 | 0 | — |
case-14 | pass→pass | 6,116 | 2,472 | -60% | 1 | 1 | 0% | 1,080 | 952 | -12% | 0 | 0 | — |
case-15 | pass→pass | 8,736 | 2,926 | -67% | 1 | 1 | 0% | 1,505 | 1,092 | -27% | 0 | 0 | — |
case-16 | fail→pass | 13,176 | 5,664 | -57% | 1 | 1 | 0% | 2,766 | 1,737 | -37% | 0 | 0 | — |
case-17 | fail→pass | 7,136 | 3,185 | -55% | 1 | 1 | 0% | 1,209 | 1,067 | -12% | 0 | 0 | — |
case-18 | fail→pass | 7,840 | 4,113 | -48% | 1 | 1 | 0% | 1,497 | 1,401 | -6% | 0 | 0 | — |
case-20 | fail→pass | 9,331 | 3,147 | -66% | 1 | 1 | 0% | 1,608 | 1,079 | -33% | 0 | 0 | — |
case-21 | pass→pass | 17,580 | 4,763 | -73% | 1 | 1 | 0% | 2,281 | 1,465 | -36% | 0 | 0 | — |
case-22 | pass→pass | 4,313 | 3,222 | -25% | 1 | 1 | 0% | 856 | 1,106 | +29% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +50 percentage points is the difference between those two pass rates over the 20 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.