Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Produces a concise, structured analysis report of a codebase: dependencies, workflow, and context.
.claude/skills/tariux-project-analyzer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 458% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -10% | 0% |
System Instruction: Project Analyzer Mode
You are a codebase and dependency analyzer. Your only function is to produce a concise, structured analysis report from the provided files, project directory, or attached code. Do not modify, refactor, or suggest any changes. Do not offer opinions, advice, or next steps.
Output exactly the following report structure. Use clear, factual language. No conversational filler.
Keep the entire report between 500 and 1000 words if possible. Prioritize precision over completeness; omit non-essential details. The report must be immediately reusable as context for future AI interactions and easily understandable by a developer.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 40,494 | 27,858 | -31% | 1 | 1 | 0% | 8,242 | 3,948 | -52% | 0 | 0 | — |
case-02 | fail→fail | 5,217 | 7,972 | +53% | 1 | 1 | 0% | 240 | 1,709 | +612% | 0 | 0 | — |
case-03 | fail→pass | 3,993 | 17,107 | +328% | 1 | 1 | 0% | 613 | 3,422 | +458% | 0 | 0 | — |
case-04 | fail→pass | 22,438 | 16,758 | -25% | 1 | 1 | 0% | 3,731 | 3,224 | -14% | 0 | 0 | — |
case-05 | fail→pass | 22,234 | 11,920 | -46% | 1 | 1 | 0% | 3,878 | 2,335 | -40% | 0 | 0 | — |
case-06 | fail→pass | 21,629 | 13,832 | -36% | 1 | 1 | 0% | 3,390 | 2,730 | -19% | 0 | 0 | — |
case-07 | fail→pass | 17,653 | 11,865 | -33% | 1 | 1 | 0% | 2,601 | 2,329 | -10% | 0 | 0 | — |
case-08 | pass→pass | 18,550 | 16,897 | -9% | 1 | 1 | 0% | 3,031 | 3,138 | +4% | 0 | 0 | — |
case-09 | fail→pass | 4,503 | 26,045 | +478% | 1 | 1 | 0% | 728 | 5,111 | +602% | 0 | 0 | — |
case-10 | fail→pass | 20,162 | 13,869 | -31% | 1 | 1 | 0% | 3,089 | 2,608 | -16% | 0 | 0 | — |
case-11 | fail→pass | 21,686 | 13,234 | -39% | 1 | 1 | 0% | 3,755 | 2,513 | -33% | 0 | 0 | — |
case-12 | fail→pass | 32,175 | 18,042 | -44% | 1 | 1 | 0% | 5,480 | 3,532 | -36% | 0 | 0 | — |
case-13 | fail→pass | 21,998 | 14,932 | -32% | 1 | 1 | 0% | 3,558 | 2,839 | -20% | 0 | 0 | — |
case-14 | fail→pass | 34,083 | 12,781 | -63% | 1 | 1 | 0% | 5,794 | 2,565 | -56% | 0 | 0 | — |
case-15 | fail→pass | 18,734 | 12,294 | -34% | 1 | 1 | 0% | 2,878 | 2,456 | -15% | 0 | 0 | — |
case-16 | fail→pass | 10,647 | 13,190 | +24% | 1 | 1 | 0% | 1,794 | 2,557 | +43% | 0 | 0 | — |
case-17 | fail→pass | 9,229 | 19,016 | +106% | 1 | 1 | 0% | 1,644 | 3,641 | +121% | 0 | 0 | — |
case-18 | fail→pass | 27,603 | 17,526 | -37% | 1 | 1 | 0% | 4,461 | 3,371 | -24% | 0 | 0 | — |
case-19 | fail→pass | 27,715 | 14,093 | -49% | 1 | 1 | 0% | 4,657 | 2,783 | -40% | 0 | 0 | — |
case-20 | fail→fail | 9,688 | 7,502 | -23% | 1 | 1 | 0% | 1,460 | 1,609 | +10% | 0 | 0 | — |
case-21 | pass→fail | 23,122 | 18,044 | -22% | 1 | 1 | 0% | 3,827 | 3,331 | -13% | 0 | 0 | — |
case-22 | fail→fail | 6,452 | 8,157 | +26% | 1 | 1 | 0% | 1,030 | 1,689 | +64% | 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 21 counted toward the lift figure. The other 1 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 +68 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is 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.