Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Verify Elixir/Phoenix changes — compile, format, and test in one loop. Use after implementation, before PRs, or after fixing bugs.
.claude/skills/oliver-kriska-verify/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 115% | 0% |
Project-aware verification for Elixir/Phoenix. Reads mix.exs and .check.exs to discover tools, test commands, and custom aliases before running anything.
mix.exs first; never run mix credo if credo isn't a dependency:ex_check + .check.exs exists, mix check replaces individual stepsmix ci or composite alias exists, use it over individual stepsRead mix.exs — extract deps/0, aliases/0, and cli/0 (for preferred_envs). Also check for .check.exs. See ${CLAUDE_SKILL_DIR}/references/project-discovery.md for full patterns.
Discover tools (deps): :credo, :dialyxir, :sobelow, :ex_check, :excoveralls, :boundary
Discover test commands (aliases + deps):
mix test (always), or custom alias like mix test.with_coveragemix playwright.test, mix cypress.run, or similar (check preferred_envs for MIX_ENV)mix playwright.run (skips setup — for re-runs)Discover composite runner: If .check.exs exists, read it — mix check may handle compile, format, credo, test, dialyzer, sobelow, and more.
Report discovery:
Project tools: compile ✓ | format ✓ | credo ✓ | dialyzer ✓ | sobelow ✓ | ex_check ✓
Test commands: mix test (unit) | mix playwright.test (E2E, MIX_ENV=int_test)
Composite runner: mix check (.check.exs covers: compiler, formatter, credo, dialyzer, sobelow, tests)
Strategy: Running `mix check` then asking about E2ECRITICAL: Before using ANY discovered alias or composite command, verify it works:
mix.lock (not just mix.exs) — deps may not be fetchedIf ex_check installed + .check.exs exists: Try mix check. If it fails, fall back to individual steps.
If composite alias found (e.g., mix ci, mix precommit): Try it. If it fails, fall back to individual steps.
Otherwise (or after fallback): Run individual steps, skipping unavailable tools.
mix compile --warnings-as-errors — always
> Elixir 1.20+ (OTP 27+): the compiler's built-in type checker emits type > violations / verified bugs as warnings, so --warnings-as-errors now fails > the build on them — no Dialyzer needed. If a previously-green build fails > after a 1.20 bump, suspect a newly-detected type violation, not a regression. > Read the message literally (accepted vs supplied type); it is almost always a > real bug. See elixir-idioms/references/elixir-120-type-system.md.
mix format --check-formatted — always (auto-fix with mix format if fails)
mix credo --strict — if :credo in deps, else skip
mix test --trace — use project test alias if available
mix dialyzer — if :dialyxir in deps, pre-PR only
mix sobelow --config — if :sobelow in deps
Skip unavailable tools with: "Credo: ⏭ Not installed"
After core verification passes, check if project has additional test commands (E2E, integration, coverage). Ask the user:
Core verification passed. Additional test commands available:
1. mix playwright.test (E2E, MIX_ENV=int_test) — ~5min
2. mix test.with_coverage (unit + coverage report)
Run any of these? [1/2/both/skip]Respect preferred_envs / cli/0 for correct MIX_ENV on each command.
| Step | Command | Condition | |------|---------|-----------| | Discovery | Read mix.exs + .check.exs | Always first | | Composite | mix check | If :ex_check installed | | Compile | mix compile --warnings-as-errors | Always | | Format | mix format --check-formatted | Always | | Credo | mix credo --strict | :credo in deps | | Test | mix test --trace | Always (use alias if exists) | | Dialyzer | mix dialyzer | :dialyxir in deps, pre-PR | | Sobelow | mix sobelow --config | :sobelow in deps | | E2E/Extra | Ask user | If additional test commands found |
/phx:verify — discovery happens automatically| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,504 | 1,596 | -75% | 1 | 1 | 0% | 1,025 | 1,615 | +58% | 0 | 0 | — |
case-02 | fail→fail | 4,548 | 2,933 | -36% | 1 | 1 | 0% | 191 | 1,684 | +782% | 0 | 0 | — |
case-03 | fail→fail | 5,192 | 4,501 | -13% | 1 | 1 | 0% | 265 | 1,631 | +515% | 0 | 0 | — |
case-04 | pass→pass | 13,682 | 9,323 | -32% | 1 | 1 | 0% | 2,583 | 3,181 | +23% | 0 | 0 | — |
case-05 | pass→fail | 4,123 | 4,874 | +18% | 1 | 1 | 0% | 757 | 1,629 | +115% | 0 | 0 | — |
case-06 | pass→fail | 7,616 | 4,784 | -37% | 1 | 1 | 0% | 1,383 | 1,645 | +19% | 0 | 0 | — |
case-07 | fail→pass | 9,395 | 4,676 | -50% | 1 | 1 | 0% | 1,677 | 2,128 | +27% | 0 | 0 | — |
case-08 | pass→pass | 1,873 | 2,833 | +51% | 1 | 1 | 0% | 277 | 1,834 | +562% | 0 | 0 | — |
case-09 | pass→fail | 3,146 | 9,229 | +193% | 1 | 1 | 0% | 472 | 2,522 | +434% | 0 | 0 | — |
case-10 | pass→fail | 3,266 | 4,669 | +43% | 1 | 1 | 0% | 474 | 1,598 | +237% | 0 | 0 | — |
case-11 | pass→fail | 4,075 | 3,790 | -7% | 1 | 1 | 0% | 560 | 1,585 | +183% | 0 | 0 | — |
case-12 | pass→fail | 4,732 | 5,539 | +17% | 1 | 1 | 0% | 746 | 1,611 | +116% | 0 | 0 | — |
case-13 | pass→pass | 8,101 | 17,349 | +114% | 1 | 1 | 0% | 1,288 | 2,507 | +95% | 0 | 0 | — |
case-14 | fail→pass | 6,344 | 3,215 | -49% | 1 | 1 | 0% | 984 | 1,787 | +82% | 0 | 0 | — |
case-15 | pass→pass | 8,621 | 4,346 | -50% | 1 | 1 | 0% | 1,597 | 2,125 | +33% | 0 | 0 | — |
case-16 | fail→pass | 10,255 | 3,371 | -67% | 1 | 1 | 0% | 1,683 | 2,036 | +21% | 0 | 0 | — |
case-17 | pass→pass | 2,443 | 3,228 | +32% | 1 | 1 | 0% | 414 | 1,947 | +370% | 0 | 0 | — |
case-18 | pass→pass | 9,628 | 3,318 | -66% | 1 | 1 | 0% | 1,599 | 1,939 | +21% | 0 | 0 | — |
case-19 | pass→pass | 8,986 | 6,370 | -29% | 1 | 1 | 0% | 1,529 | 2,519 | +65% | 0 | 0 | — |
case-20 | pass→pass | 9,847 | 5,182 | -47% | 1 | 1 | 0% | 1,662 | 2,299 | +38% | 0 | 0 | — |
case-21 | pass→pass | 2,041 | 1,448 | -29% | 1 | 1 | 0% | 311 | 1,607 | +417% | 0 | 0 | — |
case-22 | fail→pass | 7,670 | 2,443 | -68% | 1 | 1 | 0% | 1,232 | 1,724 | +40% | 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 15 counted toward the lift figure. The other 7 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 -9 percentage points is the difference between those two pass rates over the 15 comparable cases. 6 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.