Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →对进入系统的文件或 URL 执行安全与合规扫描，输出 pass/reject 判定与扫描报告，作为所有文档处理的强制前置关卡。
.claude/skills/anbeime-antinet-security-scan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 78% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 154% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -38% | 0% |
pass / reject），并写入 provenance 交由太史阁留痕。target：待检对象，二选一policy_ref：（可选）覆盖默认策略的合规规则集引用verdict：pass 或 rejectreport：安全扫描报告，含命中的黑名单项、许可状态、风险等级trace_id：本次扫描的 provenance 追踪 ID（用于太史阁留痕）python-magic）reject 并写入原因，交由管理员确认。policy_ref 适配不同行业的合规基线。scripts/run_security_scan.pycore.runtime.AgentSession.run_stage("security-scan")，调用 security.jinyiwei.JinYiWeiAgent（纯 Python，零外部依赖，可离线运行）。python skills/security-scan/scripts/run_security_scan.pyexamples/snse_survey/skill_outputs/security_scan.json（黄卡 + 扫描报告）。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-22 | fail→fail | 9,817 | 9,848 | +0% | 1 | 1 | 0% | 1,655 | 2,441 | +47% | 0 | 0 | — |
case-07 | fail→fail | 18,351 | 15,347 | -16% | 1 | 1 | 0% | 2,196 | 2,306 | +5% | 0 | 0 | — |
case-01 | fail→pass | 13,442 | 111,033 | +726% | 1 | 1 | 0% | 1,521 | 2,703 | +78% | 0 | 0 | — |
case-02 | fail→pass | 18,362 | 16,064 | -13% | 1 | 1 | 0% | 950 | 2,412 | +154% | 0 | 0 | — |
case-03 | fail→pass | 14,236 | 12,315 | -13% | 1 | 1 | 0% | 1,921 | 2,093 | +9% | 0 | 0 | — |
case-04 | pass→pass | 17,477 | 9,241 | -47% | 1 | 1 | 0% | 1,911 | 1,218 | -36% | 0 | 0 | — |
case-05 | pass→pass | 17,198 | 10,226 | -41% | 1 | 1 | 0% | 1,650 | 1,419 | -14% | 0 | 0 | — |
case-06 | fail→pass | 18,587 | 14,967 | -19% | 1 | 1 | 0% | 1,861 | 1,348 | -28% | 0 | 0 | — |
case-08 | pass→pass | 15,875 | 14,862 | -6% | 1 | 1 | 0% | 2,015 | 2,364 | +17% | 0 | 0 | — |
case-09 | fail→pass | 17,783 | 8,983 | -49% | 1 | 1 | 0% | 1,380 | 860 | -38% | 0 | 0 | — |
case-10 | fail→pass | 162,599 | 5,528 | -97% | 1 | 1 | 0% | 2,899 | 1,133 | -61% | 0 | 0 | — |
case-11 | fail→pass | 6,839 | 27,989 | +309% | 1 | 1 | 0% | 1,127 | 877 | -22% | 0 | 0 | — |
case-16 | fail→pass | 18,092 | 19,309 | +7% | 1 | 1 | 0% | 2,040 | 2,660 | +30% | 0 | 0 | — |
case-12 | fail→pass | 18,547 | 9,588 | -48% | 1 | 1 | 0% | 1,183 | 937 | -21% | 0 | 0 | — |
case-13 | fail→pass | 15,311 | 8,484 | -45% | 1 | 1 | 0% | 1,713 | 957 | -44% | 0 | 0 | — |
case-14 | pass→pass | 14,273 | 7,139 | -50% | 1 | 1 | 0% | 1,427 | 876 | -39% | 0 | 0 | — |
case-15 | fail→pass | 20,719 | 9,894 | -52% | 1 | 1 | 0% | 2,226 | 1,271 | -43% | 0 | 0 | — |
case-17 | fail→pass | 18,293 | 7,031 | -62% | 1 | 1 | 0% | 2,209 | 849 | -62% | 0 | 0 | — |
case-18 | pass→pass | 17,514 | 4,256 | -76% | 1 | 1 | 0% | 2,885 | 1,095 | -62% | 0 | 0 | — |
case-19 | pass→pass | 17,396 | 2,296 | -87% | 1 | 1 | 0% | 2,755 | 923 | -66% | 0 | 0 | — |
case-20 | pass→pass | 9,630 | 8,603 | -11% | 1 | 1 | 0% | 1,584 | 1,004 | -37% | 0 | 0 | — |
case-21 | fail→fail | 33,228 | 13,127 | -60% | 1 | 1 | 0% | 2,338 | 1,899 | -19% | 0 | 0 | — |
case-23 | pass→pass | 7,558 | 10,589 | +40% | 1 | 1 | 0% | 1,555 | 2,388 | +54% | 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. 23 cases were attempted. The headline lift of +52 percentage points is the difference between those two pass rates over the 23 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.