claude-code - 💡(How to fix) Fix [FEATURE] /insights feature suggestions should check existing settings.json and CLAUDE.md before recommending tools/hooks/MCPs that are already configured.

Preflight Checklist

• I have searched existing requests and this feature hasn't been requested yet
• This is a single feature request (not multiple features)

Problem Statement

Description:

Summary

The /insights report generates feature suggestions (MCP servers, hooks, CLAUDE.md additions) without reading the user's existing ~/.claude/settings.json or ~/.claude/CLAUDE.md, resulting in recommendations to add things that are already configured.

Steps to Reproduce

1. Have an MCP server already configured in ~/.claude/settings.json (e.g. @modelcontextprotocol/server-github)
2. Run /insights
3. Observe that the report recommends adding that same MCP server

Expected Behavior

Before generating the "Features to Try" and "CLAUDE.md Additions" sections, /insights should read:

• ~/.claude/settings.json — to check which MCP servers, hooks, and permissions are already configured
• ~/.claude/CLAUDE.md — to check which instructions are already documented
• .claude/settings.json in the current project, if present

Recommendations should be filtered or marked as already-done when the suggested configuration already exists.

Actual Behavior

The report infers what tooling the user might benefit from based solely on session transcripts, without cross-checking the actual configuration on disk. This produces duplicate recommendations that erode trust in the report's other suggestions.

Example

The report suggested: claude mcp add github -- npx -y @modelcontextprotocol/server-github

While ~/.claude/settings.json already contained: "github": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"], "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "${GITHUB_PERSONAL_ACCESS_TOKEN}" } }

Proposed Solution

At report generation time, collect a "current setup snapshot": installed MCPs, existing hooks, CLAUDE.md content, and pass it as context when building the recommendations prompt, so suggestions can be filtered against what is already in place.

Alternative Solutions

No response

Priority

Low - Nice to have

Feature Category

CLI commands and flags

Use Case Example

No response

Additional Context

No response