Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Backend TypeScript and Cloud VM development rules for cmux. Use when editing web/app/api, web/services, backend scripts, Cloud VM lifecycle, provider integrations, Postgres, Stack Auth pricing gates, migrations, or provider image build scripts.
.claude/skills/manaflow-ai-cmux-backend/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -53% | 0% |
web/app/api/**, web/services/**, and backend scripts touching providers, databases, auth, rate limits, retries, timeouts, or telemetry.cmux-prod in organization cmux. The runtime reads DATABASE_URL with CMUX_DB_DRIVER=url; migration jobs use the protected DATABASE_URL secret. AWS credentials are not database credentials.bun run cloud-vm:migrate -- staging followed by -- production; never from Vercel build or route startup. Local dev keeps the CMUX_PORT-derived Docker Postgres path from bun dev.Cloud VM build, test, and local dev scripts read provider secrets from ~/.secrets/cmux.env: FREESTYLE_API_KEY and the R2 upload vars web/scripts/build-cloud-vm-images.ts needs when creating Freestyle snapshots.
bashset -a source ~/.secrets/cmux.env set +a
~/.secrets/cmuxterm-dev.env holds local Stack/web env and not the provider build keys. bun dev sources ~/.secrets/cmux.env first when present, then ~/.secrets/cmuxterm-dev.env, so cmuxterm-specific Stack settings override broader cmux secrets. The web dev loader still accepts the legacy ~/.secret/cmuxterm.env and ~/.secrets/cmuxterm.env paths while machines migrate.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 100,843 | 19,376 | -81% | 1 | 1 | 0% | 5,476 | 4,166 | -24% | 0 | 0 | — |
case-02 | fail→pass | 23,104 | 6,982 | -70% | 1 | 1 | 0% | 3,921 | 1,681 | -57% | 0 | 0 | — |
case-03 | fail→pass | 20,737 | 7,973 | -62% | 1 | 1 | 0% | 2,980 | 1,915 | -36% | 0 | 0 | — |
case-04 | fail→pass | 18,172 | 5,050 | -72% | 1 | 1 | 0% | 2,507 | 1,290 | -49% | 0 | 0 | — |
case-05 | pass→pass | 17,253 | 4,762 | -72% | 1 | 1 | 0% | 2,149 | 1,209 | -44% | 0 | 0 | — |
case-06 | pass→pass | 10,837 | 11,312 | +4% | 1 | 1 | 0% | 1,470 | 1,210 | -18% | 0 | 0 | — |
case-07 | fail→pass | 49,636 | 5,607 | -89% | 1 | 1 | 0% | 2,806 | 1,309 | -53% | 0 | 0 | — |
case-08 | pass→pass | 19,227 | 5,809 | -70% | 1 | 1 | 0% | 2,527 | 1,332 | -47% | 0 | 0 | — |
case-09 | fail→pass | 32,021 | 3,354 | -90% | 1 | 1 | 0% | 2,312 | 1,018 | -56% | 0 | 0 | — |
case-10 | fail→pass | 16,338 | 14,530 | -11% | 1 | 1 | 0% | 2,179 | 1,041 | -52% | 0 | 0 | — |
case-11 | fail→pass | 30,918 | 7,037 | -77% | 1 | 1 | 0% | 2,620 | 1,525 | -42% | 0 | 0 | — |
case-12 | fail→pass | 28,986 | 7,050 | -76% | 1 | 1 | 0% | 2,574 | 838 | -67% | 0 | 0 | — |
case-13 | pass→pass | 15,839 | 18,360 | +16% | 1 | 1 | 0% | 1,601 | 1,071 | -33% | 0 | 0 | — |
case-14 | fail→pass | 17,040 | 3,931 | -77% | 1 | 1 | 0% | 2,881 | 1,146 | -60% | 0 | 0 | — |
case-15 | fail→pass | 11,376 | 2,838 | -75% | 1 | 1 | 0% | 1,228 | 966 | -21% | 0 | 0 | — |
case-16 | fail→pass | 18,500 | 15,927 | -14% | 1 | 1 | 0% | 3,286 | 3,391 | +3% | 0 | 0 | — |
case-17 | pass→pass | 22,945 | 11,521 | -50% | 1 | 1 | 0% | 3,337 | 2,347 | -30% | 0 | 0 | — |
case-18 | pass→pass | 17,310 | 5,922 | -66% | 1 | 1 | 0% | 2,402 | 1,249 | -48% | 0 | 0 | — |
case-19 | fail→pass | 36,092 | 5,654 | -84% | 1 | 1 | 0% | 2,546 | 1,250 | -51% | 0 | 0 | — |
case-20 | fail→pass | 17,004 | 3,705 | -78% | 1 | 1 | 0% | 2,291 | 1,058 | -54% | 0 | 0 | — |
case-21 | fail→pass | 23,326 | 31,297 | +34% | 1 | 1 | 0% | 4,010 | 3,815 | -5% | 0 | 0 | — |
case-22 | pass→pass | 9,230 | 7,450 | -19% | 1 | 1 | 0% | 1,359 | 1,081 | -20% | 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 +68 percentage points is the difference between those two pass rates over the 22 comparable cases.
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 | 9/9/2026 | +55% |
| gemini-3.6-flash | verified | 8/4/2026 | +59% |
Other measured skills in the registry, with their headline benchmark lift.