Compound could benefit from auto-setting mode based on model in Claude Code

[0xdiid]

I noticed that /ce:compound asks me whether I want to do a quick or more in-depth compound, citing context/compaction as a reason to do a quicker compound. However, with the default for Claude Max users being Opus 4.6 with the 1M context, compaction is something I rarely ever see anymore. It might benefit from either not asking at all, or citing token usage as the reasoning instead

[tmchow]

looking at the ce:compound skill, it also needed a different fix. It can't reliably do what it claims to do (detect context window pressure) so it was being conservative in practice and offering options for light versus full. Instead, I'm just always going to recommend the full but offer a lightweight as an option so you have control if you want something quicker or more context efficient. see: #528