elan6666 Byte Research
Research the questions that affect the decision; do not perform a generic market survey by default.
Approach
Focus on the decision and evidence that could change it. Use current primary or authoritative sources when facts may have changed, compare relevant alternatives when useful, and distinguish sourced facts from interpretation and uncertainty. Translate the evidence into concrete implications or recommendations.
Choose source depth and breadth according to consequence. Prefer official pages, documentation, changelogs, pricing, repositories, datasets, and direct user evidence over repeated commentary.
Do not force competitor matrices, confidence labels, or specific categories when they add little value. Cite current web claims near the claims they support.
Save research only when it will be reused, audited, or handed off. Use the
project's existing convention or a concise .byte-os/RESEARCH.md; do not require
both research and competitor files.
If direct evidence disproves a reusable prior assumption, update
.byte-os/LESSONS.md with the correction, evidence, and future verification rule.
Do not treat ordinary changes in external facts as mistakes.
Long-Running Evidence Collection
For complex research execution, default to the long-running workflow for bounded preparation, scheduled waiting when a long external collection or compute job is needed, and post-run analysis. Follow its tool-authority rules; ordinary searches do not need goal mode or an automation.
Source And Updates
Canonical repository: elan6666/your-bytedance-skills. Use its current main branch when checking for or installing updates.