Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Record a hard-won engineering lesson after a bug or mistake is resolved. Use when a bug resisted ATLAS until Boss's guidance cracked it, or when a long debugging hunt ended in an aha moment worth never relearning.
.claude/skills/syahiidkamil-learning-from-mistakes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -27% | 0% |
Capture the lesson so future-ATLAS does not pay the same debugging cost twice.
Write one — without being asked — when either is true:
Skip routine fixes. The bar is surprise: would this genuinely save someone — likely future-ATLAS — from a hard time? If the cause was obvious in hindsight to anyone, it is not worth an entry. Apply the same high-entropy filter as NOTES.md: record what is surprising, not what is expected.
One file per lesson: docs/learning-from-mistakes/<kebab-case-slug>.md. The slug names the problem, so the folder is self-indexing — e.g. stale-closure-in-useeffect.md, timezone-off-by-one-on-date-parse.md.
Keep it brief — important information only. Four short sections:
markdown# <Lesson title — the problem in one line> **Date**: YYYY-MM-DD · **Area**: <project / module / domain> ## Symptom What looked wrong — the observable behavior or the assumption that failed. ## Root cause The non-obvious truth. Why it actually happened. ## Resolution The fix, or the insight that cracked it. Credit Boss's guidance if that is what unblocked it. ## Lesson The transferable rule — how to recognize or avoid this next time.
A few tight sentences per section beats a padded template. If a section adds nothing, drop it.
Before concluding a hard bug is unsolvable, scan docs/learning-from-mistakes/ — a past lesson may already name the cause.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 13,640 | 6,200 | -55% | 1 | 1 | 0% | 1,969 | 1,493 | -24% | 0 | 0 | — |
case-02 | fail→pass | 13,414 | 6,316 | -53% | 1 | 1 | 0% | 1,986 | 1,406 | -29% | 0 | 0 | — |
case-03 | fail→pass | 10,896 | 5,764 | -47% | 1 | 1 | 0% | 1,569 | 1,384 | -12% | 0 | 0 | — |
case-04 | pass→pass | 8,145 | 2,538 | -69% | 1 | 1 | 0% | 1,229 | 836 | -32% | 0 | 0 | — |
case-05 | fail→pass | 8,234 | 3,112 | -62% | 1 | 1 | 0% | 1,312 | 900 | -31% | 0 | 0 | — |
case-06 | fail→pass | 12,341 | 5,755 | -53% | 1 | 1 | 0% | 1,943 | 1,416 | -27% | 0 | 0 | — |
case-07 | fail→pass | 11,127 | 5,702 | -49% | 1 | 1 | 0% | 1,763 | 1,361 | -23% | 0 | 0 | — |
case-08 | fail→pass | 8,703 | 6,372 | -27% | 1 | 1 | 0% | 1,309 | 1,549 | +18% | 0 | 0 | — |
case-09 | fail→pass | 17,889 | 2,724 | -85% | 1 | 1 | 0% | 2,541 | 816 | -68% | 0 | 0 | — |
case-10 | fail→pass | 13,347 | 2,982 | -78% | 1 | 1 | 0% | 2,122 | 927 | -56% | 0 | 0 | — |
case-11 | fail→pass | 17,497 | 6,644 | -62% | 1 | 1 | 0% | 2,540 | 1,461 | -42% | 0 | 0 | — |
case-12 | pass→pass | 21,728 | 17,508 | -19% | 1 | 1 | 0% | 3,400 | 3,219 | -5% | 0 | 0 | — |
case-13 | pass→pass | 19,595 | 12,437 | -37% | 1 | 1 | 0% | 3,021 | 2,238 | -26% | 0 | 0 | — |
case-14 | pass→pass | 11,264 | 10,538 | -6% | 1 | 1 | 0% | 2,039 | 2,245 | +10% | 0 | 0 | — |
case-15 | fail→pass | 17,397 | 5,546 | -68% | 1 | 1 | 0% | 2,645 | 1,332 | -50% | 0 | 0 | — |
case-16 | pass→pass | 9,946 | 4,576 | -54% | 1 | 1 | 0% | 1,526 | 1,184 | -22% | 0 | 0 | — |
case-17 | pass→pass | 10,568 | 5,004 | -53% | 1 | 1 | 0% | 1,609 | 1,109 | -31% | 0 | 0 | — |
case-18 | fail→pass | 11,434 | 3,896 | -66% | 1 | 1 | 0% | 1,693 | 1,019 | -40% | 0 | 0 | — |
case-19 | fail→pass | 12,084 | 2,706 | -78% | 1 | 1 | 0% | 1,945 | 788 | -59% | 0 | 0 | — |
case-20 | fail→pass | 17,878 | 8,037 | -55% | 1 | 1 | 0% | 2,928 | 1,659 | -43% | 0 | 0 | — |
case-21 | fail→pass | 11,167 | 3,010 | -73% | 1 | 1 | 0% | 1,811 | 968 | -47% | 0 | 0 | — |
case-22 | pass→pass | 10,424 | 3,145 | -70% | 1 | 1 | 0% | 1,593 | 893 | -44% | 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. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 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.