Vincent-20-100

Pre-commit checklist + context update — runs deterministic checks, updates PILOT.md and SESSION-CONTEXT.md. Does NOT commit.

External research with WebSearch + WebFetch + MCP — structured output to .meta/references/raw/

Unified tech-watch skill — sweep mode (automated fetch of emerging repos/articles near project themes) and deep mode (structural analysis of one specific GitHub repo). Fuses /radar (sweep) and /audit-repo (deep) with a shared card schema in .meta/references/research/.

External research with WebSearch + WebFetch + MCP — structured output to .meta/references/raw/

Fill or update the Vision section in PILOT.md — problem / target user / V1 scope / north star metric

Break work into concrete tasks with verification steps

Structured exploration before coding — one question at a time

Multi-agent adversarial debate — hybrid context (2 insiders + 1 lone wolf), structured phases, debate record. Usage: /debate <subject> [--preset X] [--deep] [--all-fresh]

Use when the user asks to lint, format, or clean up code style.

Use when the user signals end of session, asks to stop, or says save/ship/wrap up. Updates context files and prepares for clean exit.

Use when the user asks to run tests or verify the test suite.