Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use dotenvx to run commands with environment variables, manage multiple .env files, expand variables, and encrypt env files for safe commits and CI/CD.
.claude/skills/kunanonj-dotenvx/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-22 | ✓→✗ | ▼ Worse | 10% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 0% | 0% |
Use this skill when users need encrypted env workflows, multi-environment loading, or runtime env injection for any language.
npm install @dotenvx/dotenvxAlternative package managers
yarn add @dotenvx/dotenvx
pnpm add @dotenvx/dotenvx
bun add @dotenvx/dotenvxCreate a .env file in the root of your project:
ini# .env HELLO="Dotenv" OPENAI_API_KEY="your-api-key-goes-here"
Encrypt it.
dotenvx encryptAs early as possible in your application, import and configure dotenvx:
javascript// index.js require('@dotenvx/dotenvx').config() // or import '@dotenvx/dotenvx/config' // for esm console.log(`Hello ${process.env.HELLO}`)
sh$ node index.js ◇ injected env (2) from .env Hello Dotenv
That's it. process.env now has the keys and decrypted values you defined in your .env file.
.env.keys..env files as untrusted input..env artifacts in source control..env secrets should not be exposed.jsrequire('@dotenvx/dotenvx').config() // or: import '@dotenvx/dotenvx/config'
Typical requests:
Response style for agents:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,025 | 7,752 | -48% | 1 | 1 | 0% | 3,209 | 2,357 | -27% | 0 | 0 | — |
case-02 | fail→fail | 6,701 | 5,736 | -14% | 1 | 1 | 0% | 1,189 | 1,782 | +50% | 0 | 0 | — |
case-03 | pass→pass | 9,461 | 5,647 | -40% | 1 | 1 | 0% | 1,967 | 1,962 | -0% | 0 | 0 | — |
case-04 | pass→pass | 5,321 | 2,404 | -55% | 1 | 1 | 0% | 997 | 1,196 | +20% | 0 | 0 | — |
case-05 | pass→pass | 3,845 | 4,324 | +12% | 1 | 1 | 0% | 756 | 1,460 | +93% | 0 | 0 | — |
case-06 | pass→pass | 2,881 | 2,581 | -10% | 1 | 1 | 0% | 472 | 1,055 | +124% | 0 | 0 | — |
case-07 | pass→pass | 2,712 | 1,166 | -57% | 1 | 1 | 0% | 483 | 892 | +85% | 0 | 0 | — |
case-08 | pass→pass | 2,463 | 1,530 | -38% | 1 | 1 | 0% | 456 | 965 | +112% | 0 | 0 | — |
case-09 | pass→pass | 10,722 | 4,403 | -59% | 1 | 1 | 0% | 1,931 | 1,600 | -17% | 0 | 0 | — |
case-10 | fail→pass | 10,262 | 6,568 | -36% | 1 | 1 | 0% | 1,878 | 2,023 | +8% | 0 | 0 | — |
case-11 | pass→pass | 5,796 | 3,010 | -48% | 1 | 1 | 0% | 1,133 | 1,208 | +7% | 0 | 0 | — |
case-12 | pass→pass | 5,451 | 3,232 | -41% | 1 | 1 | 0% | 1,081 | 1,349 | +25% | 0 | 0 | — |
case-13 | pass→pass | 3,997 | 3,327 | -17% | 1 | 1 | 0% | 725 | 1,353 | +87% | 0 | 0 | — |
case-14 | pass→pass | 4,625 | 2,226 | -52% | 1 | 1 | 0% | 857 | 1,136 | +33% | 0 | 0 | — |
case-15 | pass→pass | 2,790 | 1,931 | -31% | 1 | 1 | 0% | 511 | 980 | +92% | 0 | 0 | — |
case-16 | pass→pass | 6,201 | 3,307 | -47% | 1 | 1 | 0% | 1,138 | 1,294 | +14% | 0 | 0 | — |
case-17 | pass→pass | 5,567 | 4,176 | -25% | 1 | 1 | 0% | 1,197 | 1,549 | +29% | 0 | 0 | — |
case-18 | pass→pass | 4,835 | 3,198 | -34% | 1 | 1 | 0% | 1,001 | 1,293 | +29% | 0 | 0 | — |
case-19 | pass→pass | 12,794 | 6,378 | -50% | 1 | 1 | 0% | 2,276 | 1,846 | -19% | 0 | 0 | — |
case-20 | fail→pass | 10,883 | 5,038 | -54% | 1 | 1 | 0% | 1,928 | 1,688 | -12% | 0 | 0 | — |
case-21 | pass→pass | 7,724 | 6,033 | -22% | 1 | 1 | 0% | 1,477 | 1,612 | +9% | 0 | 0 | — |
case-22 | pass→fail | 7,600 | 5,122 | -33% | 1 | 1 | 0% | 1,494 | 1,642 | +10% | 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 +9 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.