Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Persistent read-only advisor mission. Build a mental map of the repository once, then answer the executor's questions concisely across many turns without re-reading everything.
.claude/skills/th0rgal-orchestrator-advisor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 741% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 174% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-14 | ✓→✗ | ▼ Worse | -65% | 0% |
| case-16 | ✓→✗ | ▼ Worse | 123% | 0% |
You are a persistent advisor: a strong-reasoning agent that a cheaper executor agent consults when it hits a dead end. You live for the whole mission — each question arrives as a new user message in the same session, so your accumulated understanding of the repository is your main value.
On your first message, invest in context once:
layout of the areas the first question touches.
Do not repeat this exploration on later questions — only read files that the new question specifically requires.
git commit/push, no installs, no deletes). You advise; the executor implements.
lines, cite exact file paths (and line numbers when useful), give the specific change or diagnosis rather than general guidance.
wrong path, say so explicitly in your first sentence, then give the correct direction.
verification step beats confident guessing.
delivers the answer to the executor. Do not start open-ended background work.
looking for sibling cases in the same root-cause family. Return one bounded ledger after the family is explored; do not ask the executor to push after every example.
command to verify).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-14 | pass→fail | 14,041 | 4,503 | -68% | 1 | 1 | 0% | 1,979 | 697 | -65% | 0 | 0 | — |
case-01 | fail→fail | 14,675 | 31,186 | +113% | 1 | 1 | 0% | 2,360 | 749 | -68% | 0 | 0 | — |
case-02 | fail→fail | 32,760 | 4,069 | -88% | 1 | 1 | 0% | 1,634 | 811 | -50% | 0 | 0 | — |
case-03 | fail→fail | 28,429 | 5,644 | -80% | 1 | 1 | 0% | 1,927 | 735 | -62% | 0 | 0 | — |
case-04 | fail→pass | 7,318 | 14,010 | +91% | 1 | 1 | 0% | 213 | 1,792 | +741% | 0 | 0 | — |
case-05 | pass→pass | 3,150 | 5,871 | +86% | 1 | 1 | 0% | 340 | 1,344 | +295% | 0 | 0 | — |
case-06 | fail→pass | 3,721 | 8,510 | +129% | 1 | 1 | 0% | 402 | 1,102 | +174% | 0 | 0 | — |
case-07 | fail→fail | 11,535 | 4,286 | -63% | 1 | 1 | 0% | 1,818 | 750 | -59% | 0 | 0 | — |
case-08 | fail→fail | 8,379 | 4,425 | -47% | 1 | 1 | 0% | 1,200 | 650 | -46% | 0 | 0 | — |
case-09 | fail→fail | 9,262 | 4,099 | -56% | 1 | 1 | 0% | 1,368 | 719 | -47% | 0 | 0 | — |
case-10 | fail→fail | 10,473 | 3,363 | -68% | 1 | 1 | 0% | 1,502 | 661 | -56% | 0 | 0 | — |
case-11 | fail→fail | 10,386 | 3,793 | -63% | 1 | 1 | 0% | 1,690 | 700 | -59% | 0 | 0 | — |
case-12 | fail→fail | 8,201 | 7,021 | -14% | 1 | 1 | 0% | 1,266 | 1,527 | +21% | 0 | 0 | — |
case-13 | fail→fail | 9,244 | 5,430 | -41% | 1 | 1 | 0% | 1,839 | 666 | -64% | 0 | 0 | — |
case-15 | fail→fail | 4,730 | 5,142 | +9% | 1 | 1 | 0% | 739 | 797 | +8% | 0 | 0 | — |
case-16 | pass→fail | 1,734 | 3,658 | +111% | 1 | 1 | 0% | 278 | 620 | +123% | 0 | 0 | — |
case-17 | fail→fail | 12,287 | 4,670 | -62% | 1 | 1 | 0% | 2,026 | 702 | -65% | 0 | 0 | — |
case-18 | fail→fail | 16,831 | 2,657 | -84% | 1 | 1 | 0% | 2,559 | 624 | -76% | 0 | 0 | — |
case-19 | fail→fail | 5,219 | 3,864 | -26% | 1 | 1 | 0% | 793 | 632 | -20% | 0 | 0 | — |
case-20 | fail→pass | 9,751 | 2,752 | -72% | 1 | 1 | 0% | 1,588 | 837 | -47% | 0 | 0 | — |
case-21 | fail→fail | 8,646 | 13,310 | +54% | 1 | 1 | 0% | 1,332 | 724 | -46% | 0 | 0 | — |
case-22 | fail→fail | 7,159 | 5,624 | -21% | 1 | 1 | 0% | 1,113 | 695 | -38% | 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 4 counted toward the lift figure. The other 18 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 +5 percentage points is the difference between those two pass rates over the 4 comparable cases. 4 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.