Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create optimized, secure, production-ready Dockerfiles based on user requirements and application context.
.claude/skills/abdullahkhawer-create-dockerfile/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -47% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -31% | 0% |
Run this skill when asked to create or improve a Dockerfile for an application.
Ask the user for all details about their Docker/container requirements. Cover at minimum:
Also scan the current working directory for existing dependency files to infer requirements automatically:
bashls
Read any relevant files found (e.g. package.json, requirements.txt, pom.xml, go.mod, Gemfile, Cargo.toml) before composing the Dockerfile.
Write the Dockerfile to the working directory following these mandatory practices:
RUN layer they are created.alpine, slim, distroless) unless the user specifies otherwise.latest..env files, or secrets into the image..dockerignore note if one is missing — advise the user to create it.COPY instead of ADD unless extracting archives.WORKDIR explicitly.CMD in exec form (["executable", "arg"]), not shell form, for proper signal handling.EXPOSE to document the port (informational only).After writing the file, provide a brief explanation of:
.dockerignore, scanning with docker scout or trivy).| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | pass→pass | 9,150 | 9,044 | -1% | 1 | 1 | 0% | 1,972 | 2,565 | +30% | 0 | 0 | — |
case-01 | fail→fail | 2,445 | 3,955 | +62% | 1 | 1 | 0% | 253 | 789 | +212% | 0 | 0 | — |
case-02 | fail→fail | 18,910 | 25,715 | +36% | 1 | 1 | 0% | 3,925 | 829 | -79% | 0 | 0 | — |
case-03 | fail→fail | 15,847 | 2,780 | -82% | 1 | 1 | 0% | 165 | 821 | +398% | 0 | 0 | — |
case-04 | pass→fail | 10,241 | 4,705 | -54% | 1 | 1 | 0% | 1,930 | 1,017 | -47% | 0 | 0 | — |
case-05 | fail→pass | 9,325 | 13,384 | +44% | 1 | 1 | 0% | 1,706 | 3,112 | +82% | 0 | 0 | — |
case-06 | fail→fail | 4,296 | 4,806 | +12% | 1 | 1 | 0% | 877 | 1,007 | +15% | 0 | 0 | — |
case-07 | fail→fail | 11,590 | 5,970 | -48% | 1 | 1 | 0% | 1,885 | 916 | -51% | 0 | 0 | — |
case-08 | pass→fail | 6,532 | 4,214 | -35% | 1 | 1 | 0% | 1,353 | 931 | -31% | 0 | 0 | — |
case-09 | pass→fail | 5,325 | 7,909 | +49% | 1 | 1 | 0% | 999 | 842 | -16% | 0 | 0 | — |
case-10 | fail→fail | 6,596 | 5,327 | -19% | 1 | 1 | 0% | 1,216 | 966 | -21% | 0 | 0 | — |
case-11 | fail→fail | 7,899 | 3,867 | -51% | 1 | 1 | 0% | 1,434 | 920 | -36% | 0 | 0 | — |
case-12 | fail→pass | 9,136 | 14,666 | +61% | 1 | 1 | 0% | 1,654 | 2,894 | +75% | 0 | 0 | — |
case-13 | pass→fail | 15,034 | 3,070 | -80% | 1 | 1 | 0% | 3,238 | 862 | -73% | 0 | 0 | — |
case-14 | fail→fail | 15,001 | 6,486 | -57% | 1 | 1 | 0% | 1,361 | 1,995 | +47% | 0 | 0 | — |
case-21 | pass→pass | 3,188 | 4,574 | +43% | 1 | 1 | 0% | 743 | 1,597 | +115% | 0 | 0 | — |
case-15 | fail→fail | 12,015 | 2,455 | -80% | 1 | 1 | 0% | 2,309 | 803 | -65% | 0 | 0 | — |
case-16 | fail→pass | 9,292 | 13,697 | +47% | 1 | 1 | 0% | 1,557 | 2,426 | +56% | 0 | 0 | — |
case-17 | pass→fail | 11,524 | 2,332 | -80% | 1 | 1 | 0% | 2,265 | 797 | -65% | 0 | 0 | — |
case-18 | fail→fail | 14,332 | 6,715 | -53% | 1 | 1 | 0% | 1,918 | 848 | -56% | 0 | 0 | — |
case-19 | pass→pass | 6,890 | 13,459 | +95% | 1 | 1 | 0% | 1,305 | 2,944 | +126% | 0 | 0 | — |
case-22 | pass→fail | 10,064 | 5,862 | -42% | 1 | 1 | 0% | 2,121 | 987 | -53% | 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 7 counted toward the lift figure. The other 15 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 -14 percentage points is the difference between those two pass rates over the 7 comparable cases. 8 cases got worse with the skill loaded, and they are 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.