Suggestion: claude-cognition-skills for long-horizon collaboration quality

[hegu-1]

I'd like to suggest a skill set that targets a different layer of Claude Code usage: long-horizon collaboration quality.

Repo:

https://github.com/hegu-1/claude-cognition-skills

This is not a UI, coding, or documentation workflow skill. It focuses on failure modes that appear when Claude Code is used across long sessions and multi-step work:

• session drift from the original problem into unnecessary refactors
• premature closure before the problem structure is grounded
• high output momentum while path quality degrades
• summaries / handoffs that preserve facts but lose decision logic
• translation of founder/product judgment into generic polished language

The repo currently includes six Claude Code skills:

• nepm-kernel
• nepm-orchestrator
• trap-detector
• helpful-now-deriver
• decision-snapshot-builder
• founder-thought-translator

Possible category:

Claude Code Operations / Workflow Methodology / Long-horizon Collaboration

Suggested one-liner:

Skills for detecting path drift, premature closure, and judgment loss in long-horizon Claude Code collaboration.

The repo is newly public, so I understand if it is too early for inclusion. I am opening this issue mainly to surface the category: cognitive infrastructure for long-running AI collaboration, not just task execution.