Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Backend engineering judgment, distilled from a stronger model - invoke when CHOOSING a tech stack, language, database, queue, or architecture; designing a service, API, business logic, or schema; making a system production-ready (observability, failure handling, security); or reviewing server-side code and judging codebase health. Scenario-driven stack tradeoffs, logic-design rules, data discipline, production floors, and a rot catalog (the early signs of unmaintainable code).
.claude/skills/telagod-backend/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -59% | 0% |
Rule content lives in the five files below; this SKILL.md only routes (doctrine/04-maintenance.md governs edits to this bundle too).
| You are about to… | Read (in this folder) | |---|---| | Pick a language / framework / DB / queue / architecture, or judge a stack proposal | stack.md | | Design or review a service, module, API, or piece of business logic | logic.md | | Design a schema, transaction, migration, or cache | data.md | | Ship to production — observability, config, failure modes, security, jobs, limits | operate.md | | Review existing code, plan a refactor, or explain why a codebase feels dangerous | rot.md |
Starting a new system usually means stack.md → logic.md → data.md, in that order: stack only after the scenario interrogation, endpoints only after the data model.
Server-side engineering judgment only. Whether to delegate and when to escalate → doctrine; how to investigate/verify anything → methods; the visual/UI layer → frontend.
The one test governing the whole bundle is logic.md's opening line — is change still cheap and safe — and rot.md's opening defines the failure mode and the ugly-but-safe tiebreak. Read those two openings before judging any backend.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,099 | 15,835 | +5% | 1 | 1 | 0% | 2,597 | 2,676 | +3% | 0 | 0 | — |
case-02 | pass→pass | 19,368 | 23,069 | +19% | 1 | 1 | 0% | 3,061 | 3,316 | +8% | 0 | 0 | — |
case-03 | fail→fail | 26,262 | 25,261 | -4% | 1 | 1 | 0% | 4,179 | 4,316 | +3% | 0 | 0 | — |
case-04 | fail→fail | 27,046 | 18,354 | -32% | 1 | 1 | 0% | 1,027 | 3,258 | +217% | 0 | 0 | — |
case-21 | pass→pass | 20,634 | 18,463 | -11% | 1 | 1 | 0% | 3,181 | 3,146 | -1% | 0 | 0 | — |
case-05 | pass→pass | 16,948 | 11,227 | -34% | 1 | 1 | 0% | 2,347 | 1,981 | -16% | 0 | 0 | — |
case-06 | fail→pass | 15,349 | 12,943 | -16% | 1 | 1 | 0% | 2,319 | 2,149 | -7% | 0 | 0 | — |
case-07 | fail→pass | 14,926 | 7,593 | -49% | 1 | 1 | 0% | 2,078 | 1,453 | -30% | 0 | 0 | — |
case-08 | pass→pass | 16,105 | 10,556 | -34% | 1 | 1 | 0% | 2,498 | 1,974 | -21% | 0 | 0 | — |
case-09 | pass→pass | 15,570 | 13,464 | -14% | 1 | 1 | 0% | 2,260 | 2,149 | -5% | 0 | 0 | — |
case-10 | pass→pass | 14,299 | 6,814 | -52% | 1 | 1 | 0% | 2,251 | 1,513 | -33% | 0 | 0 | — |
case-11 | pass→pass | 19,921 | 20,358 | +2% | 1 | 1 | 0% | 2,873 | 3,364 | +17% | 0 | 0 | — |
case-12 | fail→pass | 12,819 | 4,693 | -63% | 1 | 1 | 0% | 2,051 | 1,183 | -42% | 0 | 0 | — |
case-22 | pass→pass | 21,321 | 20,924 | -2% | 1 | 1 | 0% | 3,420 | 3,680 | +8% | 0 | 0 | — |
case-13 | pass→pass | 13,027 | 9,653 | -26% | 1 | 1 | 0% | 1,836 | 1,725 | -6% | 0 | 0 | — |
case-14 | pass→pass | 15,357 | 13,963 | -9% | 1 | 1 | 0% | 2,445 | 2,433 | -0% | 0 | 0 | — |
case-15 | fail→pass | 10,494 | 2,358 | -78% | 1 | 1 | 0% | 1,641 | 670 | -59% | 0 | 0 | — |
case-16 | pass→pass | 8,566 | 6,322 | -26% | 1 | 1 | 0% | 1,341 | 1,233 | -8% | 0 | 0 | — |
case-17 | fail→pass | 11,415 | 10,689 | -6% | 1 | 1 | 0% | 1,770 | 1,889 | +7% | 0 | 0 | — |
case-18 | fail→pass | 15,608 | 13,856 | -11% | 1 | 1 | 0% | 2,422 | 2,402 | -1% | 0 | 0 | — |
case-19 | pass→pass | 13,353 | 9,819 | -26% | 1 | 1 | 0% | 1,925 | 1,721 | -11% | 0 | 0 | — |
case-20 | pass→pass | 19,716 | 19,039 | -3% | 1 | 1 | 0% | 3,676 | 4,110 | +12% | 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 21 counted toward the lift figure. The other 1 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 +32 percentage points is the difference between those two pass rates over the 21 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.
Other measured skills in the registry, with their headline benchmark lift.