Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Package session state for the next agent, or rehydrate it at start.
.claude/skills/notque-session-handoff/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 125% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 5% | 0% |
Two modes sharing one state contract. HANDOFF packages current work so the next agent (or future session) resumes without re-discovery. PICKUP rehydrates from that package at session start. Pick the mode from intent: ending or pausing work = HANDOFF; starting on existing work = PICKUP.
Produce a concise bullet package with these sections, in order. Every claim comes from a command you ran this session — re-run when stale, because a wrong handoff costs more than no handoff.
git status -sb summary; note local commits not yet pushed and whether you are in a worktree (pwd contains .claude/worktrees/).gh pr checks <num> when a PR exists).ps auxww | grep -E '<your-process>', plus a copy-paste attach or log-tail command (tail -f <logfile>, jobs -l). Redact secrets in command lines.Gate: every live process has a copy-paste command; every pending step is ordered. Output format: bullet list, Dense-Complete — short enough to paste into a PR comment or session note.
Rehydrate in this order, then act.
git status -sb; confirm branch, local commits, worktree path.gh pr view <num> --comments (derive the PR from the branch when unnumbered); note failing checks.Gate: branch, PR state, and first action are confirmed before any edit. Report discrepancies between the handoff and observed state; observed state wins.
skills/meta/do/references/worktree-rules.md): verify CWD, feature branch first.<redacted>.Cause: job exited between sessions. Solution: check its log file for exit status, record the finding, drop the attach step.
Cause: another agent or the user moved the worktree. Solution: report the difference and proceed from observed state.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,941 | 5,515 | -57% | 1 | 1 | 0% | 1,945 | 1,000 | -49% | 0 | 0 | — |
case-02 | fail→fail | 10,171 | 5,459 | -46% | 1 | 1 | 0% | 1,626 | 1,110 | -32% | 0 | 0 | — |
case-03 | fail→fail | 3,946 | 6,649 | +68% | 1 | 1 | 0% | 159 | 1,093 | +587% | 0 | 0 | — |
case-04 | fail→pass | 5,342 | 9,640 | +80% | 1 | 1 | 0% | 853 | 1,920 | +125% | 0 | 0 | — |
case-05 | pass→pass | 5,294 | 7,417 | +40% | 1 | 1 | 0% | 871 | 2,068 | +137% | 0 | 0 | — |
case-06 | pass→fail | 6,453 | 4,802 | -26% | 1 | 1 | 0% | 1,112 | 1,047 | -6% | 0 | 0 | — |
case-07 | fail→fail | 9,337 | 5,760 | -38% | 1 | 1 | 0% | 1,650 | 1,065 | -35% | 0 | 0 | — |
case-08 | pass→pass | 11,067 | 4,414 | -60% | 1 | 1 | 0% | 1,751 | 1,388 | -21% | 0 | 0 | — |
case-09 | pass→fail | 11,996 | 7,498 | -37% | 1 | 1 | 0% | 1,959 | 1,275 | -35% | 0 | 0 | — |
case-10 | fail→pass | 5,908 | 5,330 | -10% | 1 | 1 | 0% | 905 | 1,677 | +85% | 0 | 0 | — |
case-11 | fail→fail | 11,574 | 5,897 | -49% | 1 | 1 | 0% | 1,765 | 1,129 | -36% | 0 | 0 | — |
case-12 | pass→pass | 11,641 | 3,582 | -69% | 1 | 1 | 0% | 1,907 | 1,283 | -33% | 0 | 0 | — |
case-13 | fail→pass | 10,462 | 2,440 | -77% | 1 | 1 | 0% | 1,587 | 1,098 | -31% | 0 | 0 | — |
case-14 | fail→pass | 10,661 | 5,505 | -48% | 1 | 1 | 0% | 1,709 | 1,652 | -3% | 0 | 0 | — |
case-15 | pass→fail | 7,934 | 5,610 | -29% | 1 | 1 | 0% | 1,199 | 1,150 | -4% | 0 | 0 | — |
case-16 | fail→pass | 8,761 | 4,731 | -46% | 1 | 1 | 0% | 1,367 | 1,435 | +5% | 0 | 0 | — |
case-17 | pass→pass | 4,277 | 3,794 | -11% | 1 | 1 | 0% | 692 | 1,298 | +88% | 0 | 0 | — |
case-18 | fail→pass | 8,456 | 3,272 | -61% | 1 | 1 | 0% | 1,276 | 1,262 | -1% | 0 | 0 | — |
case-19 | fail→pass | 5,962 | 7,515 | +26% | 1 | 1 | 0% | 901 | 1,937 | +115% | 0 | 0 | — |
case-20 | pass→pass | 5,278 | 8,471 | +60% | 1 | 1 | 0% | 827 | 1,595 | +93% | 0 | 0 | — |
case-21 | pass→pass | 16,984 | 14,448 | -15% | 1 | 1 | 0% | 3,107 | 3,174 | +2% | 0 | 0 | — |
case-22 | pass→pass | 10,500 | 12,061 | +15% | 1 | 1 | 0% | 1,800 | 2,806 | +56% | 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 15 counted toward the lift figure. The other 7 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 15 comparable cases. 3 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.