task-decompositionlisted

# Task Decomposition > **CRITICAL:** NEVER generate multiple artifacts without stopping. After EACH artifact: list controversial points, explain simply, WAIT for user decision. Only then proceed. Decompose tech-spec Implementation Tasks into individual task files with parallel creation and validation. Before starting, read [quick-ref-task-decomposition.md](../quick-learning/references/quick-ref-task-decomposition.md) — top reasoning patterns for this skill (if file exists and non-empty). **Input:** `work/{feature}/tech-spec.md` (status: approved) **Output:** `work/{feature}/tasks/*.md` (validated) **Language:** Task files in English, communication in Russian ## Phase 0: Scope Estimation Before creating tasks, present the user a structural plan: 1. Read tech-spec Implementation Tasks section. 2. Estimate total lines of code for the feature. 3. **Break down into blocks of ~1200 lines (±300)**, each block into **steps of ~300 lines (±100)**. 4. Present the plan as a table: blocks → steps with line estimates. Add a **Complexity** column: - **L1** (trivial) — estimated_loc < 50 AND task description contains none of: auth, login, password, token, session, input validation, upload, file input, database query, SQL, API endpoint, CORS, RBAC, permission, PII, personal data, export fields - **Standard** — everything else L1 tasks get `reviewers: [code-reviewer]` only (security-auditor and test-reviewer skipped). Audit Wave covers security holistically at feature level r