Уточнение намерения перед запуском агента

research

vc-intent-clarify is a Claude Code skill that clarifies user intent before RIPER-5 phase delegation by scoring request ambiguity across 4 signals and generating structured, option-rich questions instead of open-ended prompts. It operates in two modes: SIMPLE (up to 5 file reads, runs inline) and DEEP (spawns a research subagent first, then populates questions with real file paths and codebase findings). The ambiguity score drives tier selection — 0–1 routes silently, 2 shows a routing summary for confirmation, 3+ triggers a full clarification suite. Six auto-skip conditions (continuation phrases, active phase execution, trivial fixes, explicit mode commands, and others) force Tier 0 with only a one-sentence restatement logged. Designed for autonomous /goal execution workflows where every routing decision requires an auditable intent record.