Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Interpret splunk-inspector findings and translate Splunk retention, RBAC, audit, search ACL, and auth posture into compliance evidence and remediation.
.claude/skills/grcengclub-splunk-inspector-expert/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -54% | 0% |
Use this skill when reviewing splunk-inspector output or planning Splunk logging and access-control remediation.
Findings are written to:
text~/.cache/claude-grc/findings/splunk-inspector/<run_id>.json
Resource types:
splunk_deploymentsplunk_indexsplunk_roleLOG-05: log retentionLOG-08: audit event coverageIAC-07: role and saved-search access controlIAC-04: SSO / authentication method visibilityadmin_all_objects, edit_roles, and indexes_edit require owner review._audit data and export it where long-term retention is required.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,541 | 29,306 | +348% | 1 | 1 | 0% | 1,004 | 2,230 | +122% | 0 | 0 | — |
case-02 | fail→fail | 19,204 | 3,861 | -80% | 1 | 1 | 0% | 3,070 | 615 | -80% | 0 | 0 | — |
case-03 | pass→pass | 14,155 | 11,281 | -20% | 1 | 1 | 0% | 2,264 | 1,989 | -12% | 0 | 0 | — |
case-04 | pass→pass | 18,663 | 22,938 | +23% | 1 | 1 | 0% | 3,230 | 4,135 | +28% | 0 | 0 | — |
case-05 | pass→fail | 17,521 | 14,915 | -15% | 1 | 1 | 0% | 2,802 | 2,623 | -6% | 0 | 0 | — |
case-06 | pass→pass | 10,652 | 4,510 | -58% | 1 | 1 | 0% | 1,719 | 1,079 | -37% | 0 | 0 | — |
case-07 | fail→fail | 12,943 | 8,216 | -37% | 1 | 1 | 0% | 2,077 | 1,633 | -21% | 0 | 0 | — |
case-08 | fail→pass | 13,454 | 7,797 | -42% | 1 | 1 | 0% | 2,083 | 1,485 | -29% | 0 | 0 | — |
case-09 | fail→pass | 12,932 | 10,102 | -22% | 1 | 1 | 0% | 1,864 | 1,867 | +0% | 0 | 0 | — |
case-10 | pass→pass | 10,147 | 8,786 | -13% | 1 | 1 | 0% | 1,532 | 1,582 | +3% | 0 | 0 | — |
case-11 | pass→pass | 13,492 | 10,728 | -20% | 1 | 1 | 0% | 2,054 | 1,864 | -9% | 0 | 0 | — |
case-12 | fail→pass | 10,879 | 2,225 | -80% | 1 | 1 | 0% | 1,709 | 595 | -65% | 0 | 0 | — |
case-13 | fail→pass | 10,179 | 2,758 | -73% | 1 | 1 | 0% | 1,623 | 743 | -54% | 0 | 0 | — |
case-14 | fail→pass | 10,023 | 3,171 | -68% | 1 | 1 | 0% | 1,706 | 780 | -54% | 0 | 0 | — |
case-15 | fail→pass | 14,369 | 2,142 | -85% | 1 | 1 | 0% | 2,342 | 632 | -73% | 0 | 0 | — |
case-16 | fail→pass | 14,459 | 1,705 | -88% | 1 | 1 | 0% | 2,267 | 520 | -77% | 0 | 0 | — |
case-17 | pass→pass | 14,543 | 2,917 | -80% | 1 | 1 | 0% | 2,086 | 506 | -76% | 0 | 0 | — |
case-18 | fail→pass | 16,644 | 1,591 | -90% | 1 | 1 | 0% | 2,571 | 500 | -81% | 0 | 0 | — |
case-19 | pass→pass | 13,456 | 11,071 | -18% | 1 | 1 | 0% | 1,993 | 1,895 | -5% | 0 | 0 | — |
case-20 | pass→pass | 13,086 | 5,095 | -61% | 1 | 1 | 0% | 1,971 | 1,069 | -46% | 0 | 0 | — |
case-21 | pass→pass | 15,655 | 13,578 | -13% | 1 | 1 | 0% | 2,243 | 2,414 | +8% | 0 | 0 | — |
case-22 | pass→pass | 8,600 | 7,786 | -9% | 1 | 1 | 0% | 1,303 | 1,513 | +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. The headline lift of +32 percentage points is the difference between those two pass rates over the 22 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.