Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Entry point for Genie operations — routes bug reports, questions, and operational commands, resumes existing lifecycle state, and orchestrates work that needs durable planning or coordination. Other ordinary requests bypass the lifecycle with a one-line notice unless the user asks for Genie.
.claude/skills/automagik-dev-genie/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -24% | 0% |
Genie is the execution system and the source of task truth. When a workspace contains .genie/, treat Genie state — wishes, tasks, boards, worker status — as canonical over anything scraped from a terminal.
skills directly: wish, work, and review. This plugin no longer ships genie-work/genie-review duplicates — the product skills are canonical.
genie_board,genie_wish_status, genie_task) rather than terminal scraping. The native Hermes surface only adds the gap tools MCP does not cover: genie_status (doctor plus .genie presence), genie_work_plan (context --plan spawn preview), and genie_review_plan (wish status plus acceptance criteria).
tmux capture-pane or sleep loops to infer workerprogress. The structured tools return the same truth with provenance.
mutation: "none").genie spawn, a non-plan genie context resolution, genie taskdone — require explicit human approval before they run, every time.
the genie ... argv that produced the data, on one line.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,383 | 4,436 | -31% | 1 | 1 | 0% | 338 | 695 | +106% | 0 | 0 | — |
case-02 | fail→fail | 3,999 | 6,209 | +55% | 1 | 1 | 0% | 424 | 757 | +79% | 0 | 0 | — |
case-03 | fail→fail | 5,515 | 6,197 | +12% | 1 | 1 | 0% | 280 | 788 | +181% | 0 | 0 | — |
case-04 | fail→pass | 16,343 | 4,769 | -71% | 1 | 1 | 0% | 2,344 | 1,157 | -51% | 0 | 0 | — |
case-05 | fail→pass | 13,493 | 2,596 | -81% | 1 | 1 | 0% | 2,027 | 765 | -62% | 0 | 0 | — |
case-06 | fail→pass | 5,764 | 5,591 | -3% | 1 | 1 | 0% | 844 | 800 | -5% | 0 | 0 | — |
case-07 | fail→pass | 8,648 | 2,755 | -68% | 1 | 1 | 0% | 1,300 | 718 | -45% | 0 | 0 | — |
case-08 | fail→pass | 7,228 | 3,501 | -52% | 1 | 1 | 0% | 1,194 | 902 | -24% | 0 | 0 | — |
case-09 | fail→pass | 5,801 | 2,459 | -58% | 1 | 1 | 0% | 787 | 761 | -3% | 0 | 0 | — |
case-10 | fail→pass | 5,999 | 4,295 | -28% | 1 | 1 | 0% | 866 | 1,054 | +22% | 0 | 0 | — |
case-11 | pass→pass | 7,638 | 2,578 | -66% | 1 | 1 | 0% | 1,054 | 808 | -23% | 0 | 0 | — |
case-12 | pass→pass | 6,630 | 3,190 | -52% | 1 | 1 | 0% | 964 | 842 | -13% | 0 | 0 | — |
case-13 | fail→fail | 6,859 | 1,881 | -73% | 1 | 1 | 0% | 1,032 | 678 | -34% | 0 | 0 | — |
case-14 | pass→pass | 8,661 | 1,892 | -78% | 1 | 1 | 0% | 1,232 | 636 | -48% | 0 | 0 | — |
case-15 | fail→pass | 8,026 | 2,359 | -71% | 1 | 1 | 0% | 1,408 | 696 | -51% | 0 | 0 | — |
case-16 | fail→pass | 7,557 | 2,816 | -63% | 1 | 1 | 0% | 1,127 | 815 | -28% | 0 | 0 | — |
case-17 | fail→pass | 4,978 | 2,700 | -46% | 1 | 1 | 0% | 790 | 750 | -5% | 0 | 0 | — |
case-18 | pass→pass | 7,085 | 2,943 | -58% | 1 | 1 | 0% | 1,041 | 875 | -16% | 0 | 0 | — |
case-19 | pass→pass | 9,733 | 2,057 | -79% | 1 | 1 | 0% | 1,379 | 680 | -51% | 0 | 0 | — |
case-20 | pass→pass | 8,577 | 5,471 | -36% | 1 | 1 | 0% | 1,487 | 1,237 | -17% | 0 | 0 | — |
case-21 | pass→pass | 9,013 | 4,743 | -47% | 1 | 1 | 0% | 1,592 | 1,143 | -28% | 0 | 0 | — |
case-22 | pass→pass | 10,829 | 5,629 | -48% | 1 | 1 | 0% | 1,982 | 1,408 | -29% | 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 19 counted toward the lift figure. The other 3 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 +45 percentage points is the difference between those two pass rates over the 19 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.