Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Manually route a query to the optimal Claude model (Haiku/Sonnet/Opus)
.claude/skills/aiskillstore-route/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 184% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 403% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 269% | 0% |
Override automatic model selection and force a specific Claude model for your query.
/route <model> <query>Where <model> is one of:
haiku or fast - Use Haiku for simple, quick taskssonnet or standard - Use Sonnet for typical coding tasksopus or deep - Use Opus for complex analysisCRITICAL: The user's explicit model choice MUST be honored. Do NOT override, reclassify, or second-guess their selection. This is a MANUAL OVERRIDE command - the entire point is to bypass automatic classification.
Parse $ARGUMENTS to extract the model and query:
DO NOT:
The user said which model. Use that model. Period.
| Argument | Executor | Model | |----------|----------|-------| | haiku or fast | fast-executor | Haiku | | sonnet or standard | standard-executor | Sonnet | | opus or deep | deep-executor | Opus |
/route opus What's the syntax for a TypeScript interface?Result: Routes to Opus (deep-executor) regardless of query complexity.
/route haiku Fix the authentication bug in login.tsResult: Routes to Haiku (fast-executor) for cost savings.
/route sonnet Design a caching systemResult: Routes to Sonnet (standard-executor).
If the user doesn't provide a valid model, respond with:
Usage: /route <model> <query>
Models:
haiku, fast - Quick, simple tasks (cheapest)
sonnet, standard - Typical coding tasks (default)
opus, deep - Complex analysis (most capable)
Example: /route opus Analyze the security of this authentication flow| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,891 | 7,988 | -60% | 1 | 1 | 0% | 1,209 | 1,051 | -13% | 0 | 0 | — |
case-02 | fail→fail | 7,748 | 13,832 | +79% | 1 | 1 | 0% | 1,514 | 2,564 | +69% | 0 | 0 | — |
case-03 | fail→fail | 5,650 | 6,941 | +23% | 1 | 1 | 0% | 943 | 1,722 | +83% | 0 | 0 | — |
case-04 | fail→fail | 4,293 | 3,624 | -16% | 1 | 1 | 0% | 609 | 1,161 | +91% | 0 | 0 | — |
case-05 | fail→fail | 1,997 | 13,083 | +555% | 1 | 1 | 0% | 279 | 2,251 | +707% | 0 | 0 | — |
case-06 | fail→fail | 17,910 | 15,509 | -13% | 1 | 1 | 0% | 3,565 | 3,496 | -2% | 0 | 0 | — |
case-07 | fail→pass | 3,179 | 1,902 | -40% | 1 | 1 | 0% | 337 | 958 | +184% | 0 | 0 | — |
case-08 | fail→pass | 4,951 | 1,556 | -69% | 1 | 1 | 0% | 767 | 847 | +10% | 0 | 0 | — |
case-09 | fail→pass | 2,178 | 2,697 | +24% | 1 | 1 | 0% | 208 | 1,047 | +403% | 0 | 0 | — |
case-10 | fail→fail | 2,122 | 9,544 | +350% | 1 | 1 | 0% | 203 | 1,300 | +540% | 0 | 0 | — |
case-11 | fail→fail | 4,443 | 9,081 | +104% | 1 | 1 | 0% | 688 | 1,931 | +181% | 0 | 0 | — |
case-12 | fail→fail | 2,694 | 7,257 | +169% | 1 | 1 | 0% | 294 | 905 | +208% | 0 | 0 | — |
case-13 | fail→pass | 6,483 | 2,228 | -66% | 1 | 1 | 0% | 997 | 965 | -3% | 0 | 0 | — |
case-14 | fail→pass | 2,433 | 2,654 | +9% | 1 | 1 | 0% | 254 | 937 | +269% | 0 | 0 | — |
case-15 | fail→fail | 16,193 | 39,010 | +141% | 1 | 1 | 0% | 2,518 | 3,416 | +36% | 0 | 0 | — |
case-20 | pass→pass | 14,430 | 10,109 | -30% | 1 | 1 | 0% | 2,268 | 2,075 | -9% | 0 | 0 | — |
case-16 | fail→fail | 15,781 | 10,562 | -33% | 1 | 1 | 0% | 2,574 | 2,500 | -3% | 0 | 0 | — |
case-17 | fail→fail | 11,851 | 8,626 | -27% | 1 | 1 | 0% | 2,269 | 2,199 | -3% | 0 | 0 | — |
case-18 | pass→pass | 2,054 | 11,485 | +459% | 1 | 1 | 0% | 247 | 2,795 | +1032% | 0 | 0 | — |
case-19 | fail→fail | 5,151 | 7,681 | +49% | 1 | 1 | 0% | 804 | 994 | +24% | 0 | 0 | — |
case-21 | pass→pass | 11,927 | 9,162 | -23% | 1 | 1 | 0% | 1,785 | 1,961 | +10% | 0 | 0 | — |
case-22 | pass→pass | 18,727 | 14,222 | -24% | 1 | 1 | 0% | 3,468 | 2,972 | -14% | 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 18 counted toward the lift figure. The other 4 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 +23 percentage points is the difference between those two pass rates over the 18 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.