Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →> Note: This skill has been adapted from claude-skillz
.claude/skills/majiayu000-switch-persona/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | 176% | 0% |
| case-12 | ✓→✗ | ▼ Worse | 187% | 0% |
| case-17 | ✓→✗ | ▼ Worse | 81% | 0% |
| case-22 | ✓→✗ | ▼ Worse | -56% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | 20% | 0% |
name: Switch Persona description: Quick persona switching. Triggers: 'switch persona', 'switch to X', 'become X'. Lists personas, reads selected file, switches immediately.
> Note: This skill has been adapted from claude-skillz > for use with GitHub Copilot Agent Skills.
User says:
→ Execute protocol below
Execute immediately:
1. Read ~/.github/system-prompts/[name].txt (or .md)
2. Adopt new persona instructions
3. Continue conversationAnnounce:
Switched to [name]. [First line of persona description]Then respond as new persona.
Step 1: List available
bashls ~/.github/system-prompts/
Present clean list:
Available personas:
1. super-tdd-developer (current)
2. requirements-expert
3. claude-code-optimizer
...
Which persona? (number or name)Step 2: Get selection
Wait for user input.
Step 3: Switch
Read selected file:
Read ~/.github/system-prompts/[selected].txtAnnounce:
Switched to [name]. [First line of persona description]Then respond as new persona.
When switching:
File not found:
Persona '[name]' not found. Available: [list]Read failed:
Cannot read [name]. Error: [details]Quick, simple persona switching. No confirmations, no ceremony.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | fail→fail | 8,508 | 8,429 | -1% | 1 | 1 | 0% | 465 | 943 | +103% | 0 | 0 | — |
case-20 | pass→pass | 17,375 | 17,513 | +1% | 1 | 1 | 0% | 2,248 | 2,687 | +20% | 0 | 0 | — |
case-21 | pass→pass | 16,303 | 12,917 | -21% | 1 | 1 | 0% | 1,754 | 1,783 | +2% | 0 | 0 | — |
case-01 | fail→fail | 9,946 | 16,978 | +71% | 1 | 1 | 0% | 719 | 888 | +24% | 0 | 0 | — |
case-02 | fail→fail | 7,782 | 6,940 | -11% | 1 | 1 | 0% | 1,110 | 719 | -35% | 0 | 0 | — |
case-03 | fail→fail | 12,275 | 16,835 | +37% | 1 | 1 | 0% | 1,003 | 864 | -14% | 0 | 0 | — |
case-04 | fail→fail | 9,344 | 16,854 | +80% | 1 | 1 | 0% | 668 | 841 | +26% | 0 | 0 | — |
case-05 | fail→fail | 10,412 | 12,584 | +21% | 1 | 1 | 0% | 798 | 1,884 | +136% | 0 | 0 | — |
case-06 | fail→fail | 15,409 | 11,659 | -24% | 1 | 1 | 0% | 1,610 | 778 | -52% | 0 | 0 | — |
case-07 | fail→fail | 10,644 | 8,148 | -23% | 1 | 1 | 0% | 782 | 860 | +10% | 0 | 0 | — |
case-08 | fail→fail | 13,110 | 24,028 | +83% | 1 | 1 | 0% | 1,230 | 2,229 | +81% | 0 | 0 | — |
case-09 | fail→fail | 9,618 | 16,921 | +76% | 1 | 1 | 0% | 630 | 926 | +47% | 0 | 0 | — |
case-10 | fail→pass | 7,663 | 16,168 | +111% | 1 | 1 | 0% | 304 | 840 | +176% | 0 | 0 | — |
case-11 | fail→fail | 3,173 | 14,338 | +352% | 1 | 1 | 0% | 441 | 3,148 | +614% | 0 | 0 | — |
case-12 | pass→fail | 7,480 | 12,399 | +66% | 1 | 1 | 0% | 313 | 898 | +187% | 0 | 0 | — |
case-13 | fail→fail | 9,978 | 6,633 | -34% | 1 | 1 | 0% | 680 | 863 | +27% | 0 | 0 | — |
case-14 | fail→fail | 7,706 | 18,607 | +141% | 1 | 1 | 0% | 422 | 917 | +117% | 0 | 0 | — |
case-15 | fail→fail | 2,835 | 13,415 | +373% | 1 | 1 | 0% | 404 | 1,021 | +153% | 0 | 0 | — |
case-16 | fail→fail | 4,268 | 16,308 | +282% | 1 | 1 | 0% | 558 | 809 | +45% | 0 | 0 | — |
case-17 | pass→fail | 3,746 | 12,970 | +246% | 1 | 1 | 0% | 556 | 1,006 | +81% | 0 | 0 | — |
case-18 | fail→fail | 3,274 | 18,389 | +462% | 1 | 1 | 0% | 429 | 835 | +95% | 0 | 0 | — |
case-22 | pass→fail | 10,949 | 10,834 | -1% | 1 | 1 | 0% | 1,721 | 759 | -56% | 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 8 counted toward the lift figure. The other 14 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 -9 percentage points is the difference between those two pass rates over the 8 comparable cases. 8 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.