Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use only when the user explicitly asks to run, set up, update, re-mine, or deepen Emulo from real local AI coding-session history.
.claude/skills/ohad6k-mine/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -60% | 0% |
Mine only real user-authored .jsonl sessions. Never synthesize a profile from rules files, memory, or a typed self-description.
EMULO_PY to emulo.py two directories above this skill, falling back to ./emulo.py only in a direct checkout. Confirm Python 3 exists.python "$EMULO_PY" plugin preflight, show valid sessions, post-dedupe source tokens, selected source tokens, cache hits, planned worker calls, and planned reducer calls, then wait for explicit cost approval before model work.--preview. Say this exactly before approval: Quick preview creates a starter profile from selected history, not the full profile. Never call preview the default, the full profile, or equivalent in quality to full-history mining.approval_hash. Run python "$EMULO_PY" plugin prepare --approved-plan-hash HASH with the exact approved mode. If the hash changed, show the new plan and obtain approval again. Retain the exact run_id, selected segment/report paths, and pack_path.MINING_PROMPT.md, writes its assigned JSON report, and runs the read-only plugin validate-report command until accepted.plugin cache-report; stop on rejection.plugin validate-pack.plugin activate, then run plugin status and render the card. Report active version, domain states, selected source tokens, actual worker/reducer passes, cache reuse, and targeted-deepen instructions.Every prepared mode requires explicit approval of its displayed plan. Installation itself scans no logs and schedules zero mining calls. An identical update schedules zero additional Emulo mining calls, although the host task still has normal interaction overhead.
Adaptive receipt/scout stages are experimental and excluded from the Plugin release path. Run --stage A only when a developer explicitly requests experimental adaptive-recall testing; never select it automatically or use it for release calibration.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-18 | fail→pass | 13,053 | 2,572 | -80% | 1 | 1 | 0% | 1,707 | 848 | -50% | 0 | 0 | — |
case-07 | fail→pass | 6,546 | 3,840 | -41% | 1 | 1 | 0% | 859 | 1,228 | +43% | 0 | 0 | — |
case-06 | pass→pass | 10,677 | 10,954 | +3% | 1 | 1 | 0% | 2,010 | 2,677 | +33% | 0 | 0 | — |
case-01 | fail→fail | 26,482 | 12,810 | -52% | 1 | 1 | 0% | 1,558 | 1,041 | -33% | 0 | 0 | — |
case-02 | fail→fail | 7,316 | 8,904 | +22% | 1 | 1 | 0% | 1,108 | 986 | -11% | 0 | 0 | — |
case-03 | fail→fail | 6,059 | 10,497 | +73% | 1 | 1 | 0% | 837 | 1,537 | +84% | 0 | 0 | — |
case-04 | pass→fail | 9,795 | 12,562 | +28% | 1 | 1 | 0% | 1,904 | 2,815 | +48% | 0 | 0 | — |
case-05 | pass→pass | 16,393 | 17,124 | +4% | 1 | 1 | 0% | 3,623 | 4,440 | +23% | 0 | 0 | — |
case-08 | fail→pass | 17,226 | 2,743 | -84% | 1 | 1 | 0% | 2,461 | 908 | -63% | 0 | 0 | — |
case-09 | fail→pass | 7,217 | 3,239 | -55% | 1 | 1 | 0% | 915 | 1,063 | +16% | 0 | 0 | — |
case-10 | fail→fail | 12,147 | 4,136 | -66% | 1 | 1 | 0% | 1,736 | 1,376 | -21% | 0 | 0 | — |
case-11 | fail→pass | 14,548 | 3,168 | -78% | 1 | 1 | 0% | 2,429 | 976 | -60% | 0 | 0 | — |
case-12 | fail→fail | 19,502 | 2,749 | -86% | 1 | 1 | 0% | 2,559 | 899 | -65% | 0 | 0 | — |
case-13 | fail→pass | 21,548 | 2,094 | -90% | 1 | 1 | 0% | 2,967 | 842 | -72% | 0 | 0 | — |
case-14 | fail→pass | 14,067 | 5,342 | -62% | 1 | 1 | 0% | 1,957 | 1,263 | -35% | 0 | 0 | — |
case-15 | pass→pass | 15,535 | 33,358 | +115% | 1 | 1 | 0% | 2,196 | 909 | -59% | 0 | 0 | — |
case-16 | fail→pass | 28,177 | 2,629 | -91% | 1 | 1 | 0% | 3,820 | 844 | -78% | 0 | 0 | — |
case-17 | pass→pass | 11,576 | 3,614 | -69% | 1 | 1 | 0% | 1,574 | 1,015 | -36% | 0 | 0 | — |
case-19 | fail→pass | 12,910 | 3,078 | -76% | 1 | 1 | 0% | 1,725 | 963 | -44% | 0 | 0 | — |
case-20 | pass→pass | 8,877 | 3,151 | -65% | 1 | 1 | 0% | 1,058 | 951 | -10% | 0 | 0 | — |
case-21 | fail→pass | 11,917 | 3,315 | -72% | 1 | 1 | 0% | 1,883 | 934 | -50% | 0 | 0 | — |
case-22 | pass→pass | 15,728 | 5,330 | -66% | 1 | 1 | 0% | 2,016 | 1,301 | -35% | 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 +41 percentage points is the difference between those two pass rates over the 19 comparable cases. 2 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.