skills/walkinglabs__learn-harness-engineering

mirror of https://github.com/walkinglabs/learn-harness-engineering.git synced 2026-09-19 05:06:27 +08:00

Files

T

devtulkin 414b3e471a docs(uz): add Uzbek translation of full course

Adds a complete Uzbek (Latin script) translation under docs/uz/,
following the existing multilingual pattern (en/zh/ko/vi/ru).
Discussed with @sanbuphy in advance — contribution welcomed.

Scope:
- 12 lectures (full prose + code-sample docs)
- 6 projects (overview pages)
- Resources: templates, reference, openai-advanced pack
- Skills overview
- VitePress config: registered uz locale (nav, sidebar, outline labels)

Translation discipline:
- Uzbek Latin orthography normalized (oʻ/gʻ U+02BB, ʼ U+02BC,
  curly quotes “…” in prose, straight " inside HTML/mermaid syntax)
- Project-wide glossary (docs/uz/GLOSSARY.md) — ~80 technical terms
  with consistent renderings; explicit anti-calque list for
  common idiom traps
- Style guide (docs/uz/STYLE-GUIDE.md) — orthography rules,
  capitalisation conventions, common spelling pitfalls
- Helper scripts under scripts/ to enforce orthography and
  spot-check terminology drift

All 12 lectures went through translate → manual review →
LQA review cycles with extensive native-speaker feedback applied.

Note: feedback is welcome on terminology choices, especially
for Uzbek renderings of newer terms (sprint contract, evaluator
rubric, isometric model control, etc.).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-11 01:01:42 +03:00

2.0 KiB

Raw Permalink Blame History

Loyihalarga xush kelibsiz

Bu Learn Harness Engineering kursining amaliyot qismidir. Faqat maʼruzalarni oʻqish yetarli emas — muhitlarni oʻzingiz qurishingiz va turli qoidalar ostida Codex, Claude Code yoki boshqa AI agentlar qanday harakat qilishini kuzatishingiz kerak.

Loyiha xulosasi

Ushbu kurs 6 ta bosqichma-bosqich, amaliy loyihalarni oʻz ichiga oladi va ular sizga noldan boshlab qanday qilib ishonchli agent ishlash muhitini qurishni oʻrgatadi:

Faqat prompt vs. Avval qoidalar (Prompt-Only vs. Rules-First): Agent faqat prompt bilan qanday ishlashini va minimal harness bilan qanday ishlashini taqqoslang.
Agent oʻqiy oladigan ish maydoni (Agent-Readable Workspace): Repozitoriyni qanday qilib AI uchun qulay qilishni va ishlarni topshirish (handoff) mexanizmlarini yaratishni oʻrganing.
Koʻp sessiyali uzluksizlik (Multi-Session Continuity): Agent ishlarni sessiyalar oʻrtasida muammosiz davom ettira olishi uchun holat fayllari va ishga tushirish (initialization) skriptlarini loyihalang.
Runtime qayta aloqa va skoup nazorati (Runtime Feedback and Scope Control): Agentga oʻz kodini oʻzi test qilish va ishlash davomida (runtime) xatolarni tuzatish imkonini beruvchi vositalarni joriy qiling.
Oʻz-oʻzini tekshirish va rollarni ajratish (Self-Verification and Role Separation): Gallyutsinatsiyalar va vaqtidan oldin gʻalabani eʼlon qilishning oldini olish uchun mustaqil tekshiruv mexanizmini quring.
Toʻliq harness (Capstone): Yakuniy, kuzatiladigan (observable), end-to-end agent ish muhitini toʻplang.

Qanday davom ettirish kerak

Har bir loyiha papkasi odatda quyidagilarni oʻz ichiga oladi:

starter/: Sizning boshlangʻich ish maydoningiz (workspace).
solution/: Namuna yechim (agar qiyin vaziyatga tushib qolsangiz).
Sizning orqa foningiz va aniq maqsadlaringiz tushuntirilgan vazifa yoʻriqnomalari.

starter/ katalogidagi vazifalarni bajarish uchun oʻzingiz afzal koʻrgan AI kod yozish agentidan (masalan, Claude Code, Cursor, Trae) foydalaning.

2.0 KiB Raw Permalink Blame History Unescape Escape

Loyihalarga xush kelibsiz

Loyiha xulosasi

Qanday davom ettirish kerak

2.0 KiB

Raw Permalink Blame History