[FEATURE] /insights should acknowledge incomplete behavioral data or ground analysis with external sources beyond session-only activity

Resolved 💬 2 comments Opened Feb 8, 2026 by charlieCollins Closed Mar 9, 2026

Preflight Checklist

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

Problem Statement

The insights report draws behavioral conclusions from incomplete data — it only sees activity within Claude Code sessions, not the user's broader workflow. For example, it flags "only 2 commits across 49 hours" and infers I'm doing "exploratory work that never gets validated," when in reality I'm making dozens of commits from a separate terminal window throughout every session.

The report treats Claude Code as if it's the entire development environment, but for many users it's one tool alongside their own terminal, editor, git workflow, IDEs, other LLMs, and may other things. Suggestions and analysis built on this incomplete picture end up feeling inaccurate and undermine trust in the report.

Proposed Solution

The report should either acknowledge its limited visibility explicitly, or could do things such as (one example) look at the actual git log for the repo to get ground-truth commit activity regardless of where it originated.

Alternative Solutions

_No response_

Priority

Low - Nice to have

Feature Category

Other

Use Case Example

_No response_

Additional Context

_No response_

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗