Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Capture an out-of-scope improvement opportunity so it doesn't get lost. Use when the user asks to "note improvement", "save improvement", "track this for later", "remember this improvement", "note this idea", "log improvement", "backlog this", or "park this idea". Also invoke proactively when noticing something improvable during work that falls outside the current task's scope, or after deliberately shipping a simpler approach that accepts a known ceiling — briefly mention it to the user and off
.claude/skills/tobihagemann-note-improvement/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 1478% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 245% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-12 | ✓→✗ | ▼ Worse | 33% | 0% |
| case-16 | ✓→✗ | ▼ Worse | 37% | 0% |
Capture improvement opportunities discovered during work so they don't get silently dropped. Appends to a project-level .turbo/improvements.md file that serves as a backlog of actionable ideas.
Find the nearest .git directory or project root. The improvements file lives at .turbo/improvements.md relative to the project root.
Gather from context or $ARGUMENTS:
direct, investigate, or plan — see criteria belowrefactor, performance, reliability, readability, testing, docs, dx (developer experience), or feature$implement.When the criteria above clearly select one value, use it. Otherwise, use request_user_input to confirm; default to plan if the user declines to choose.
Read .turbo/improvements.md if it exists. Create it with the header below if it doesn't.
File header (only when creating new):
markdown# Improvements Out-of-scope improvement opportunities captured during work sessions. Review periodically and pull items into active work when appropriate.
Entry format:
markdown### <one-line summary> - **Type**: <direct | investigate | plan> - **Category**: <category> - **Where**: `<file path or area>` - **Why**: <brief rationale> - **Ceiling**: <limit the shipped approach accepts> - **Revisit**: <condition that makes the fuller version worth building> - **Noted**: <YYYY-MM-DD>
Include the Ceiling and Revisit lines when the entry records a deliberate simplification; omit both otherwise. Append the new entry at the end of the file.
Tell the user the improvement was noted and where the file is.
trivial or standard, rewrite the Type to direct or plan respectively so the file converges on current vocabulary..turbo/ directory, use request_user_input to confirm the location before creating one.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,524 | 7,637 | +17% | 1 | 1 | 0% | 1,032 | 1,004 | -3% | 0 | 0 | — |
case-02 | fail→fail | 7,437 | 6,723 | -10% | 1 | 1 | 0% | 1,026 | 1,308 | +27% | 0 | 0 | — |
case-03 | fail→fail | 3,932 | 5,083 | +29% | 1 | 1 | 0% | 578 | 963 | +67% | 0 | 0 | — |
case-04 | fail→fail | 6,165 | 10,197 | +65% | 1 | 1 | 0% | 972 | 2,534 | +161% | 0 | 0 | — |
case-05 | fail→pass | 1,944 | 17,334 | +792% | 1 | 1 | 0% | 239 | 3,771 | +1478% | 0 | 0 | — |
case-06 | fail→fail | 10,667 | 10,054 | -6% | 1 | 1 | 0% | 1,715 | 2,433 | +42% | 0 | 0 | — |
case-07 | fail→fail | 4,371 | 10,837 | +148% | 1 | 1 | 0% | 710 | 1,615 | +127% | 0 | 0 | — |
case-08 | fail→fail | 5,110 | 9,165 | +79% | 1 | 1 | 0% | 930 | 1,069 | +15% | 0 | 0 | — |
case-13 | fail→pass | 5,337 | 11,940 | +124% | 1 | 1 | 0% | 794 | 2,743 | +245% | 0 | 0 | — |
case-09 | fail→fail | 8,475 | 6,258 | -26% | 1 | 1 | 0% | 827 | 1,107 | +34% | 0 | 0 | — |
case-10 | fail→fail | 8,569 | 8,618 | +1% | 1 | 1 | 0% | 1,356 | 1,170 | -14% | 0 | 0 | — |
case-11 | fail→fail | 4,382 | 5,316 | +21% | 1 | 1 | 0% | 678 | 947 | +40% | 0 | 0 | — |
case-12 | pass→fail | 5,122 | 6,459 | +26% | 1 | 1 | 0% | 812 | 1,081 | +33% | 0 | 0 | — |
case-14 | fail→fail | 7,650 | 9,062 | +18% | 1 | 1 | 0% | 1,158 | 988 | -15% | 0 | 0 | — |
case-15 | fail→fail | 4,740 | 6,777 | +43% | 1 | 1 | 0% | 243 | 1,178 | +385% | 0 | 0 | — |
case-16 | pass→fail | 5,894 | 6,515 | +11% | 1 | 1 | 0% | 949 | 1,296 | +37% | 0 | 0 | — |
case-17 | fail→fail | 4,427 | 4,309 | -3% | 1 | 1 | 0% | 717 | 962 | +34% | 0 | 0 | — |
case-18 | fail→fail | 9,522 | 7,473 | -22% | 1 | 1 | 0% | 1,789 | 1,075 | -40% | 0 | 0 | — |
case-19 | fail→fail | 7,083 | 6,531 | -8% | 1 | 1 | 0% | 1,059 | 1,181 | +12% | 0 | 0 | — |
case-20 | pass→fail | 5,519 | 3,584 | -35% | 1 | 1 | 0% | 897 | 950 | +6% | 0 | 0 | — |
case-21 | fail→pass | 9,693 | 15,308 | +58% | 1 | 1 | 0% | 1,913 | 3,114 | +63% | 0 | 0 | — |
case-22 | fail→fail | 4,823 | 5,131 | +6% | 1 | 1 | 0% | 748 | 1,277 | +71% | 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 5 counted toward the lift figure. The other 17 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 0 percentage points is the difference between those two pass rates over the 5 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.