Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Mandatory iPolloWork code-change gate for modern, minimal, performant, reuse-first implementation and clean repository ownership. Use whenever AI creates, edits, deletes, or refactors application code, server code, packages, scripts, tests, dependencies, schemas, routes, UI, or generated-file workflows. Reuse existing code before creating files, keep one source of truth, prevent parallel implementations and junk directories, justify every new file or dependency, and audit the current change befo
.claude/skills/devin-axis-ipollowork-maintainable-code/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | 68% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 48% | 0% |
Treat this as the repository's single code-quality Skill. Implement changes with the smallest coherent diff, modern platform patterns, bounded runtime cost, and one clear owner. Search and reuse before creating anything.
AGENTS.md, the owning package manifest, the real entrypoint, and nearby implementations.references/repository-boundaries.md before adding a file, directory, dependency, API, table, or cross-package import.rg.apps/app/src/components/ui, apps/app/src/components, the target domain, apps/server/src, packages/types, and packages/ui as relevant.old, new, v2, fallback, or dead compatibility copies behind.powershellnode .codex/skills/ipollowork-maintainable-code/scripts/audit-changes.mjs
new, old, v2, copy, backup, temp, tmp, misc, notes, drafts, handoff, or tdd-summary paths to avoid integrating with the current owner.| Situation | Action | | --- | --- | | Existing API/component/helper meets the need | Import and reuse it | | Existing implementation differs only by presentation or configuration | Add typed props/options without changing existing defaults | | Two real call sites need the same non-trivial behavior | Extract to the nearest shared owner | | Similar-looking code has different domain rules or is unlikely to be reused | Keep it local; do not force an abstraction | | No compatible implementation exists after searching | Create one in the narrowest correct owner |
Do not copy a component and rename it, create *V2, *New, *Copy, or duplicate a helper to avoid understanding its API. Do not add a generic abstraction for a single trivial use.
any, broad type assertions, duplicated derived state, prop-to-state mirroring, hidden global state, and fallback branches that types or control flow make impossible.apps/app/src/components/ui.apps/app/src/components.apps/app/src/react-app/domains/<domain>.packages/ui only when more than one application genuinely consumes it.packages/types; do not maintain separate client and server copies.apps/server/src/routes; put reusable business behavior in the existing owning service/extension module.apps/server/src/paths.ts, including safe workspace-relative normalization and root containment helpers.path.join or string concatenation.apps/server/src or another source directory.<workspace>/design/<session-id>/... for design/PPT/web sessions and <workspace>/video/<session-id>/... for video sessions.apps/desktop; do not leak shell-specific behavior into app domains.apps/orchestrator; do not create a second orchestration path in the UI or server.iPolloWork and ipollowork; do not introduce alternate product spellings in code, paths, docs, or user-facing text.AGENTS.md.git diff --check, and the maintainability audit pass; any unverified runtime surface is stated explicitly.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,782 | 15,543 | +13% | 1 | 1 | 0% | 172 | 2,381 | +1284% | 0 | 0 | — |
case-02 | fail→fail | 32,227 | 16,253 | -50% | 1 | 1 | 0% | 5,762 | 2,489 | -57% | 0 | 0 | — |
case-03 | pass→fail | 18,466 | 16,876 | -9% | 1 | 1 | 0% | 2,709 | 2,441 | -10% | 0 | 0 | — |
case-04 | pass→pass | 16,370 | 10,138 | -38% | 1 | 1 | 0% | 1,696 | 3,029 | +79% | 0 | 0 | — |
case-05 | pass→pass | 12,377 | 10,305 | -17% | 1 | 1 | 0% | 1,392 | 3,089 | +122% | 0 | 0 | — |
case-06 | pass→pass | 14,809 | 27,972 | +89% | 1 | 1 | 0% | 1,737 | 3,448 | +99% | 0 | 0 | — |
case-07 | pass→fail | 19,118 | 17,895 | -6% | 1 | 1 | 0% | 2,328 | 2,636 | +13% | 0 | 0 | — |
case-08 | fail→fail | 21,090 | 16,631 | -21% | 1 | 1 | 0% | 2,492 | 2,572 | +3% | 0 | 0 | — |
case-09 | fail→fail | 16,831 | 9,770 | -42% | 1 | 1 | 0% | 1,757 | 2,845 | +62% | 0 | 0 | — |
case-10 | fail→pass | 16,720 | 12,200 | -27% | 1 | 1 | 0% | 1,957 | 3,295 | +68% | 0 | 0 | — |
case-11 | fail→fail | 21,280 | 16,330 | -23% | 1 | 1 | 0% | 2,791 | 2,355 | -16% | 0 | 0 | — |
case-12 | pass→pass | 18,370 | 10,726 | -42% | 1 | 1 | 0% | 1,850 | 3,011 | +63% | 0 | 0 | — |
case-13 | fail→pass | 16,540 | 4,095 | -75% | 1 | 1 | 0% | 1,812 | 2,877 | +59% | 0 | 0 | — |
case-14 | fail→fail | 16,283 | 6,974 | -57% | 1 | 1 | 0% | 1,899 | 2,548 | +34% | 0 | 0 | — |
case-15 | pass→pass | 10,612 | 3,996 | -62% | 1 | 1 | 0% | 1,548 | 2,799 | +81% | 0 | 0 | — |
case-16 | fail→pass | 13,386 | 4,961 | -63% | 1 | 1 | 0% | 2,235 | 3,056 | +37% | 0 | 0 | — |
case-17 | fail→fail | 13,846 | 8,359 | -40% | 1 | 1 | 0% | 1,697 | 2,520 | +48% | 0 | 0 | — |
case-18 | pass→pass | 12,847 | 4,363 | -66% | 1 | 1 | 0% | 2,066 | 2,917 | +41% | 0 | 0 | — |
case-19 | fail→pass | 10,587 | 5,176 | -51% | 1 | 1 | 0% | 1,627 | 2,956 | +82% | 0 | 0 | — |
case-20 | fail→pass | 12,178 | 4,817 | -60% | 1 | 1 | 0% | 2,036 | 3,012 | +48% | 0 | 0 | — |
case-21 | fail→pass | 5,863 | 4,555 | -22% | 1 | 1 | 0% | 1,018 | 2,943 | +189% | 0 | 0 | — |
case-22 | pass→pass | 11,867 | 4,343 | -63% | 1 | 1 | 0% | 1,734 | 2,784 | +61% | 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 14 counted toward the lift figure. The other 8 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 +18 percentage points is the difference between those two pass rates over the 14 comparable cases. 5 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.