Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Wrap up the current session by writing a short HANDOFF.md in the project folder, so the next session can resume quickly. Use when the user says done, wrap up, close this out, or capture the session.
.claude/skills/chrisblattman-done/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 1099% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 122% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 99% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -54% | 0% |
| case-10 | ✓→✗ | ▼ Worse | -58% | 0% |
Close out this session. Plain Write tool only — no scripts, no state files, no routing logic. Everything stays in the current folder.
./HANDOFF.md — it's a snapshot, not a log, so replace the whole file. 30 lines maximum, exactly these sections: # Handoff — <project>
## Status ## Next steps
## Key files ## Decisions
A few short lines per section. Next steps are checkboxes, most important first; leave finished items out. Key files = the handful of files the next session should open first.
./SESSION_LOG.md exists, also append a dated entry of 3–6 bullets — what happened this session. Append at the end; never edit earlier entries. (If the file doesn't exist, skip this — don't create it.)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→fail | 9,213 | 5,860 | -36% | 1 | 1 | 0% | 1,574 | 722 | -54% | 0 | 0 | — |
case-01 | fail→fail | 4,748 | 6,480 | +36% | 1 | 1 | 0% | 173 | 656 | +279% | 0 | 0 | — |
case-02 | fail→fail | 8,207 | 3,583 | -56% | 1 | 1 | 0% | 1,215 | 520 | -57% | 0 | 0 | — |
case-03 | fail→fail | 9,340 | 4,871 | -48% | 1 | 1 | 0% | 1,385 | 592 | -57% | 0 | 0 | — |
case-04 | fail→fail | 4,388 | 6,633 | +51% | 1 | 1 | 0% | 247 | 705 | +185% | 0 | 0 | — |
case-06 | fail→pass | 3,987 | 13,208 | +231% | 1 | 1 | 0% | 232 | 2,781 | +1099% | 0 | 0 | — |
case-07 | fail→pass | 7,542 | 11,469 | +52% | 1 | 1 | 0% | 1,050 | 2,327 | +122% | 0 | 0 | — |
case-08 | fail→fail | 2,938 | 7,062 | +140% | 1 | 1 | 0% | 461 | 632 | +37% | 0 | 0 | — |
case-09 | fail→fail | 6,598 | 5,487 | -17% | 1 | 1 | 0% | 981 | 643 | -34% | 0 | 0 | — |
case-10 | pass→fail | 9,907 | 7,424 | -25% | 1 | 1 | 0% | 1,583 | 660 | -58% | 0 | 0 | — |
case-21 | fail→fail | 10,656 | 15,772 | +48% | 1 | 1 | 0% | 1,572 | 3,121 | +99% | 0 | 0 | — |
case-11 | pass→fail | 5,911 | 5,752 | -3% | 1 | 1 | 0% | 866 | 608 | -30% | 0 | 0 | — |
case-12 | pass→fail | 6,928 | 11,275 | +63% | 1 | 1 | 0% | 1,061 | 694 | -35% | 0 | 0 | — |
case-13 | fail→fail | 3,067 | 7,237 | +136% | 1 | 1 | 0% | 256 | 686 | +168% | 0 | 0 | — |
case-14 | fail→pass | 9,023 | 19,253 | +113% | 1 | 1 | 0% | 1,470 | 2,924 | +99% | 0 | 0 | — |
case-15 | fail→fail | 7,462 | 4,737 | -37% | 1 | 1 | 0% | 1,121 | 699 | -38% | 0 | 0 | — |
case-16 | pass→fail | 8,994 | 12,855 | +43% | 1 | 1 | 0% | 1,441 | 1,390 | -4% | 0 | 0 | — |
case-17 | fail→fail | 9,686 | 13,849 | +43% | 1 | 1 | 0% | 1,408 | 1,911 | +36% | 0 | 0 | — |
case-18 | pass→pass | 13,707 | 11,270 | -18% | 1 | 1 | 0% | 2,172 | 1,828 | -16% | 0 | 0 | — |
case-19 | pass→fail | 7,428 | 8,138 | +10% | 1 | 1 | 0% | 1,185 | 609 | -49% | 0 | 0 | — |
case-20 | fail→fail | 7,008 | 3,759 | -46% | 1 | 1 | 0% | 1,174 | 914 | -22% | 0 | 0 | — |
case-22 | pass→fail | 6,389 | 16,762 | +162% | 1 | 1 | 0% | 968 | 3,493 | +261% | 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 6 counted toward the lift figure. The other 16 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 -18 percentage points is the difference between those two pass rates over the 6 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.