Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Spawn or manage voice teammates. Each teammate is a separate Claude personality in its own tmux pane.
.claude/skills/majiayu000-teammate/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -61% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -72% | 0% |
/claude-talk:teammate [personality] — spawn a specific personality, or omit for interactive selection.
Spawn a teammate by personality name (use Bash):
bashsource ~/.claude-talk/venvs/wlk/bin/activate claude-talk teammate spawn "$PERSONALITY"
Where $PERSONALITY is the argument passed to this skill (e.g., claude, vex, hank). If no argument, list available personalities first:
bashsource ~/.claude-talk/venvs/wlk/bin/activate claude-talk personality list
Then ask the user which personality to spawn and run claude-talk teammate spawn <chosen>.
Report back: which personality was spawned, which tmux pane they're in, and that voice routing is now active for them.
Also mention: teammates can message each other using claude-talk message send <personality> "text" or broadcast to all with claude-talk message broadcast "text". Messages arrive as Teammate <name> said: ... in the recipient's session.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-14 | fail→fail | 15,364 | 3,816 | -75% | 1 | 1 | 0% | 1,589 | 937 | -41% | 0 | 0 | — |
case-01 | fail→fail | 11,845 | 9,574 | -19% | 1 | 1 | 0% | 1,228 | 456 | -63% | 0 | 0 | — |
case-02 | pass→fail | 6,686 | 10,090 | +51% | 1 | 1 | 0% | 1,022 | 510 | -50% | 0 | 0 | — |
case-03 | fail→fail | 17,079 | 15,354 | -10% | 1 | 1 | 0% | 284 | 491 | +73% | 0 | 0 | — |
case-04 | fail→fail | 23,342 | 10,322 | -56% | 1 | 1 | 0% | 1,538 | 585 | -62% | 0 | 0 | — |
case-05 | fail→fail | 13,246 | 7,879 | -41% | 1 | 1 | 0% | 2,047 | 612 | -70% | 0 | 0 | — |
case-06 | fail→pass | 32,936 | 9,100 | -72% | 1 | 1 | 0% | 2,435 | 1,093 | -55% | 0 | 0 | — |
case-07 | fail→pass | 30,120 | 7,512 | -75% | 1 | 1 | 0% | 1,793 | 701 | -61% | 0 | 0 | — |
case-08 | fail→pass | 15,567 | 10,160 | -35% | 1 | 1 | 0% | 2,488 | 1,107 | -56% | 0 | 0 | — |
case-09 | fail→pass | 25,425 | 3,573 | -86% | 1 | 1 | 0% | 1,785 | 812 | -55% | 0 | 0 | — |
case-10 | fail→pass | 15,270 | 7,528 | -51% | 1 | 1 | 0% | 2,392 | 662 | -72% | 0 | 0 | — |
case-11 | fail→pass | 17,554 | 3,089 | -82% | 1 | 1 | 0% | 1,717 | 729 | -58% | 0 | 0 | — |
case-12 | fail→pass | 14,261 | 9,881 | -31% | 1 | 1 | 0% | 1,516 | 1,168 | -23% | 0 | 0 | — |
case-13 | fail→fail | 11,156 | 17,849 | +60% | 1 | 1 | 0% | 1,785 | 804 | -55% | 0 | 0 | — |
case-15 | fail→pass | 13,785 | 8,980 | -35% | 1 | 1 | 0% | 1,254 | 915 | -27% | 0 | 0 | — |
case-16 | fail→pass | 17,335 | 5,199 | -70% | 1 | 1 | 0% | 1,738 | 1,272 | -27% | 0 | 0 | — |
case-17 | fail→pass | 33,123 | 11,724 | -65% | 1 | 1 | 0% | 2,252 | 972 | -57% | 0 | 0 | — |
case-18 | fail→pass | 10,457 | 7,672 | -27% | 1 | 1 | 0% | 1,591 | 711 | -55% | 0 | 0 | — |
case-19 | fail→fail | 13,222 | 7,539 | -43% | 1 | 1 | 0% | 1,179 | 658 | -44% | 0 | 0 | — |
case-20 | pass→pass | 13,108 | 8,053 | -39% | 1 | 1 | 0% | 1,353 | 1,635 | +21% | 0 | 0 | — |
case-21 | pass→pass | 17,740 | 14,081 | -21% | 1 | 1 | 0% | 2,285 | 2,126 | -7% | 0 | 0 | — |
case-22 | pass→fail | 15,281 | 6,676 | -56% | 1 | 1 | 0% | 1,737 | 511 | -71% | 0 | 0 | — |
case-23 | fail→pass | 11,467 | 7,932 | -31% | 1 | 1 | 0% | 2,012 | 708 | -65% | 0 | 0 | — |
case-24 | fail→pass | 13,232 | 8,053 | -39% | 1 | 1 | 0% | 2,034 | 764 | -62% | 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. 24 cases were attempted, and 18 counted toward the lift figure. The other 6 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 +46 percentage points is the difference between those two pass rates over the 18 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
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.