Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audits and hardens API credential handling (env vars, separation, rotation plan, least privilege, auditability). Use when integrating services or preparing production deployments where secrets must be managed safely.
.claude/skills/sundial-org-api-credentials-hygiene/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 94% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 15% | 0% |
Audits and hardens API credential handling (env vars, separation, rotation plan, least privilege, auditability).
.env template (placeholders only).Success = no secrets committed or embedded, permissions minimized, rotation steps documented, and auditability defined.
Credential map template:
textCREDENTIAL MAP - Integration: <name> - Env vars: - <VAR_NAME>: <purpose> (secret/non-secret) - Permissions/scopes: <list> - Used by: <service/workflow> - Storage: <secret manager/env var> - Rotation: <cadence> | <owner> | <procedure> - Audit: <what is logged and where>
If providing a template, output assets/dotenv-template.example with placeholders only.
Output: Env var map + plan to move to n8n credentials/env vars + rotation runbook.
Output: Two env maps + naming scheme + access boundary checklist.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 13,371 | 21,216 | +59% | 1 | 1 | 0% | 2,194 | 4,262 | +94% | 0 | 0 | — |
case-02 | pass→pass | 23,410 | 17,869 | -24% | 1 | 1 | 0% | 3,557 | 3,510 | -1% | 0 | 0 | — |
case-03 | fail→pass | 18,803 | 19,540 | +4% | 1 | 1 | 0% | 2,710 | 3,943 | +45% | 0 | 0 | — |
case-04 | pass→pass | 15,312 | 9,437 | -38% | 1 | 1 | 0% | 2,186 | 2,220 | +2% | 0 | 0 | — |
case-05 | pass→pass | 4,225 | 12,881 | +205% | 1 | 1 | 0% | 533 | 2,218 | +316% | 0 | 0 | — |
case-06 | pass→pass | 11,809 | 13,240 | +12% | 1 | 1 | 0% | 2,124 | 3,127 | +47% | 0 | 0 | — |
case-07 | pass→pass | 7,550 | 15,694 | +108% | 1 | 1 | 0% | 1,219 | 3,334 | +174% | 0 | 0 | — |
case-08 | fail→fail | 8,128 | 10,482 | +29% | 1 | 1 | 0% | 1,289 | 2,562 | +99% | 0 | 0 | — |
case-09 | fail→pass | 14,310 | 5,837 | -59% | 1 | 1 | 0% | 2,092 | 1,727 | -17% | 0 | 0 | — |
case-10 | pass→pass | 20,821 | 15,674 | -25% | 1 | 1 | 0% | 3,128 | 3,318 | +6% | 0 | 0 | — |
case-11 | pass→pass | 18,684 | 19,362 | +4% | 1 | 1 | 0% | 2,792 | 3,807 | +36% | 0 | 0 | — |
case-12 | pass→pass | 17,991 | 16,344 | -9% | 1 | 1 | 0% | 2,590 | 3,230 | +25% | 0 | 0 | — |
case-13 | pass→fail | 24,749 | 18,201 | -26% | 1 | 1 | 0% | 4,081 | 3,830 | -6% | 0 | 0 | — |
case-14 | fail→pass | 19,688 | 19,057 | -3% | 1 | 1 | 0% | 3,035 | 3,894 | +28% | 0 | 0 | — |
case-15 | fail→pass | 19,513 | 15,995 | -18% | 1 | 1 | 0% | 2,935 | 3,364 | +15% | 0 | 0 | — |
case-20 | fail→pass | 15,091 | 13,156 | -13% | 1 | 1 | 0% | 2,319 | 2,819 | +22% | 0 | 0 | — |
case-16 | pass→pass | 4,478 | 5,420 | +21% | 1 | 1 | 0% | 576 | 1,590 | +176% | 0 | 0 | — |
case-17 | pass→pass | 16,648 | 17,158 | +3% | 1 | 1 | 0% | 2,516 | 3,831 | +52% | 0 | 0 | — |
case-18 | pass→pass | 14,335 | 13,957 | -3% | 1 | 1 | 0% | 2,340 | 3,130 | +34% | 0 | 0 | — |
case-19 | pass→pass | 12,963 | 16,481 | +27% | 1 | 1 | 0% | 2,168 | 3,543 | +63% | 0 | 0 | — |
case-21 | pass→pass | 16,261 | 16,993 | +5% | 1 | 1 | 0% | 2,433 | 3,500 | +44% | 0 | 0 | — |
case-22 | fail→fail | 19,871 | 14,244 | -28% | 1 | 1 | 0% | 1,752 | 2,989 | +71% | 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 +23 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/19/2026 | +36% |
Other measured skills in the registry, with their headline benchmark lift.