Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Cluster hardened instincts (high-confidence feedback_*/discovery_* memories) into a proposed higher-level structure — a Command, Skill, or Agent. Run when many related instincts have accumulated in one domain. Part of the Instinct Engine. Do NOT use for one-off pattern capture (use /patterns) or daily journaling.
.claude/skills/bilal140202-evolve/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 1% | 0% |
When many related, high-confidence instincts pile up in one domain, that is a signal to promote them into ONE reusable structure instead of leaving them as loose memories. /evolve finds those clusters deterministically and drafts a proposal you refine.
ECC source pattern: "/evolve clusters related instincts into higher-level structures: Commands / Skills / Agents." Reimplemented clean per license-hygiene.
bashpython3 ~/.claude/skills/ai-brain-starter/scripts/instinct.py evolve
It groups every instinct by inferred domain, computes each cluster's median effective confidence, and for clusters that clear the propose bar (>= 2 instincts AND median confidence >= 0.80) writes a scaffold to <vault>/⚙️ Meta/Instinct Proposals/proposed-skill-<domain>.md. Clusters below the bar print as watch (not yet ripe).
For each proposal the script wrote, decide the structure:
A cluster that is really just 2-3 facets of an existing rule should be CONSOLIDATED into that rule, not promoted. Reject those.
For an accepted cluster, draft the real skill/command body from the member instincts (keep each instinct's Action + Evidence). Place it in the right repo per the install rules, wire its discoverability + automation + verification in the SAME session (the three-layer wiring rule), then retire or link the source memories. Delete the proposal file once adopted or rejected — proposals are scratch, not a backlog.
In a cron/--print session: run Step 1, report the PROPOSE clusters and the proposal paths, and STOP. Promotion to a real skill is a judgment call that needs a human — never auto-create skills.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,083 | 3,827 | -73% | 1 | 1 | 0% | 2,430 | 678 | -72% | 0 | 0 | — |
case-02 | fail→fail | 14,899 | 2,355 | -84% | 1 | 1 | 0% | 2,297 | 741 | -68% | 0 | 0 | — |
case-03 | fail→fail | 9,151 | 1,838 | -80% | 1 | 1 | 0% | 1,672 | 720 | -57% | 0 | 0 | — |
case-04 | fail→pass | 7,955 | 3,172 | -60% | 1 | 1 | 0% | 1,548 | 1,066 | -31% | 0 | 0 | — |
case-05 | fail→pass | 6,986 | 3,692 | -47% | 1 | 1 | 0% | 1,196 | 1,118 | -7% | 0 | 0 | — |
case-06 | fail→pass | 8,413 | 1,676 | -80% | 1 | 1 | 0% | 1,487 | 753 | -49% | 0 | 0 | — |
case-07 | pass→pass | 5,539 | 2,285 | -59% | 1 | 1 | 0% | 1,010 | 928 | -8% | 0 | 0 | — |
case-08 | fail→pass | 6,727 | 2,252 | -67% | 1 | 1 | 0% | 1,099 | 931 | -15% | 0 | 0 | — |
case-09 | fail→pass | 6,616 | 3,093 | -53% | 1 | 1 | 0% | 1,033 | 1,039 | +1% | 0 | 0 | — |
case-10 | pass→pass | 7,177 | 2,119 | -70% | 1 | 1 | 0% | 1,222 | 903 | -26% | 0 | 0 | — |
case-11 | fail→pass | 7,763 | 2,631 | -66% | 1 | 1 | 0% | 1,334 | 896 | -33% | 0 | 0 | — |
case-12 | fail→pass | 8,092 | 3,408 | -58% | 1 | 1 | 0% | 1,313 | 1,045 | -20% | 0 | 0 | — |
case-13 | pass→pass | 10,527 | 2,992 | -72% | 1 | 1 | 0% | 1,724 | 958 | -44% | 0 | 0 | — |
case-14 | fail→pass | 10,336 | 3,183 | -69% | 1 | 1 | 0% | 1,674 | 1,108 | -34% | 0 | 0 | — |
case-15 | fail→fail | 11,670 | 1,591 | -86% | 1 | 1 | 0% | 1,794 | 702 | -61% | 0 | 0 | — |
case-16 | fail→pass | 8,156 | 3,019 | -63% | 1 | 1 | 0% | 1,430 | 1,064 | -26% | 0 | 0 | — |
case-17 | fail→fail | 7,788 | 1,904 | -76% | 1 | 1 | 0% | 1,269 | 785 | -38% | 0 | 0 | — |
case-18 | fail→pass | 6,012 | 1,713 | -72% | 1 | 1 | 0% | 896 | 677 | -24% | 0 | 0 | — |
case-19 | fail→pass | 8,448 | 2,338 | -72% | 1 | 1 | 0% | 1,454 | 912 | -37% | 0 | 0 | — |
case-20 | pass→pass | 7,420 | 3,592 | -52% | 1 | 1 | 0% | 1,272 | 1,123 | -12% | 0 | 0 | — |
case-21 | fail→fail | 6,413 | 3,934 | -39% | 1 | 1 | 0% | 967 | 1,107 | +14% | 0 | 0 | — |
case-22 | fail→pass | 9,161 | 4,439 | -52% | 1 | 1 | 0% | 1,546 | 1,293 | -16% | 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 21 counted toward the lift figure. The other 1 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 +55 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is 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.