Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Behavioral rules and lifecycle event handlers for the core-tools package. Includes auto-approval of file operations targeting deep-analysis session directories.
.claude/skills/majiayu000-lifecycle-hooks/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 49% | 0% |
This skill defines behavioral rules that were originally enforced by platform lifecycle hooks. These rules describe automated behaviors that trigger at specific points in the agent workflow.
Trigger: Fires before the agent executes any action (file write, shell command, etc.) Applies when: Action matches Write|Edit|Bash
This rule auto-approves file operations targeting deep-analysis session and cache directories. It prevents permission prompts from interrupting autonomous exploration workflows.
The implementation logic is in references/auto-approve-da-session.sh. This script:
.agents/sessions/__da_live__/, .agents/sessions/exploration-cache/, .agents/sessions/da-*/)What this component does: Defines automated behavioral rules that were originally enforced by platform lifecycle hooks.
Origin: Converted from 1 lifecycle hook (PreToolUse/before_action)
Capabilities needed:
Adaptation guidance:
references/ can be executed directly if the harness supports shell-based hooks.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 14,362 | 7,526 | -48% | 1 | 1 | 0% | 1,457 | 763 | -48% | 0 | 0 | — |
case-01 | fail→pass | 21,606 | 16,477 | -24% | 1 | 1 | 0% | 3,204 | 3,194 | -0% | 0 | 0 | — |
case-02 | fail→pass | 16,307 | 20,019 | +23% | 1 | 1 | 0% | 3,367 | 3,507 | +4% | 0 | 0 | — |
case-03 | fail→pass | 17,150 | 13,671 | -20% | 1 | 1 | 0% | 3,092 | 3,137 | +1% | 0 | 0 | — |
case-04 | fail→pass | 40,015 | 7,474 | -81% | 1 | 1 | 0% | 1,267 | 1,882 | +49% | 0 | 0 | — |
case-10 | fail→pass | 41,507 | 1,616 | -96% | 1 | 1 | 0% | 7,251 | 613 | -92% | 0 | 0 | — |
case-05 | fail→pass | 29,612 | 7,894 | -73% | 1 | 1 | 0% | 1,962 | 827 | -58% | 0 | 0 | — |
case-06 | fail→pass | 13,251 | 1,697 | -87% | 1 | 1 | 0% | 2,229 | 633 | -72% | 0 | 0 | — |
case-07 | fail→pass | 20,698 | 2,414 | -88% | 1 | 1 | 0% | 2,729 | 735 | -73% | 0 | 0 | — |
case-08 | fail→pass | 22,086 | 7,216 | -67% | 1 | 1 | 0% | 3,004 | 1,649 | -45% | 0 | 0 | — |
case-11 | fail→pass | 4,694 | 6,844 | +46% | 1 | 1 | 0% | 644 | 674 | +5% | 0 | 0 | — |
case-12 | pass→pass | 22,712 | 7,976 | -65% | 1 | 1 | 0% | 3,170 | 1,767 | -44% | 0 | 0 | — |
case-13 | fail→pass | 11,280 | 3,173 | -72% | 1 | 1 | 0% | 1,862 | 972 | -48% | 0 | 0 | — |
case-14 | fail→pass | 13,105 | 8,934 | -32% | 1 | 1 | 0% | 2,426 | 1,071 | -56% | 0 | 0 | — |
case-15 | fail→pass | 13,772 | 8,160 | -41% | 1 | 1 | 0% | 2,398 | 876 | -63% | 0 | 0 | — |
case-16 | fail→pass | 21,232 | 16,400 | -23% | 1 | 1 | 0% | 2,851 | 2,331 | -18% | 0 | 0 | — |
case-17 | pass→pass | 12,735 | 7,438 | -42% | 1 | 1 | 0% | 2,074 | 743 | -64% | 0 | 0 | — |
case-18 | pass→pass | 18,880 | 3,477 | -82% | 1 | 1 | 0% | 2,433 | 1,019 | -58% | 0 | 0 | — |
case-19 | fail→pass | 9,559 | 8,251 | -14% | 1 | 1 | 0% | 1,628 | 963 | -41% | 0 | 0 | — |
case-20 | pass→pass | 36,692 | 17,696 | -52% | 1 | 1 | 0% | 3,267 | 2,806 | -14% | 0 | 0 | — |
case-21 | pass→pass | 16,478 | 15,069 | -9% | 1 | 1 | 0% | 1,915 | 2,276 | +19% | 0 | 0 | — |
case-22 | pass→pass | 20,733 | 21,735 | +5% | 1 | 1 | 0% | 4,011 | 3,555 | -11% | 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 20 counted toward the lift figure. The other 2 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 +73 percentage points is the difference between those two pass rates over the 20 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.