C:/Program Files/Git/insights suggestions should consider CLAUDE.md and project configuration

Description

The /insights report generates useful observational analysis (interaction style, friction points, what's working), but its prescriptive suggestions (CLAUDE.md additions, usage patterns, "on the horizon") don't appear to read or account for the project's existing CLAUDE.md, memory files, or custom skills. This leads to suggestions that duplicate existing instructions or contradict stated project principles.

Examples

Suggesting what's already configured:

• "Bundle documentation updates into every task" — my CLAUDE.md already says "Update STATUS.md at the end of each session" under Agentic Coding Rules.
• "Front-load environment checks" — my auto-memory file already contains full environment details (Python path, shell, OS) that load into every conversation.
• "Plan-then-build parser workflow" — I had just set up a /parse custom skill that encodes exactly this workflow.

Contradicting stated project principles:

• "Parallel Agents for Multi-Platform Parser Pipeline" with a BaseParser class — my CLAUDE.md's number 1 development principle is Gall's Law ("Build one working thing at a time") and number 2 is YAGNI. This suggestion directly violates both.
• "Test-Driven Parser Development" with pytest — my CLAUDE.md explicitly says "Don't create test files until asked to."

Expected behavior

Before generating suggestions, /insights should read the project's CLAUDE.md (and ideally memory files and skills) to:

1. Filter out suggestions for things already addressed by existing configuration
2. Avoid recommending approaches that contradict the project's stated principles
3. Tailor suggestions to the project's actual conventions rather than generic best practices

Environment

• Claude Code on Windows (Git Bash / MINGW64)
• Project has CLAUDE.md with development principles, coding standards, and agentic coding rules
• Project uses auto-memory (MEMORY.md) and a custom skill (.claude/skills/parse/)