[Feature Request] Global auto-enable: track agent sessions without per-repo configuration

[rraafffaa]

Summary

As a less-technical user (Product Manager) who uses Claude Code for rapid prototyping, I frequently spin up quick POCs — creating new repos, committing code, and moving on. I don't follow a traditional developer workflow where I cd into a project directory, set up tooling, and then start working. I often create repos on the fly via gh repo create or let Claude Code scaffold things directly.

The problem: Entire currently requires per-repo configuration (entire configure or entire enable) before it can track agent sessions. This means any work done in an unconfigured repo is invisible to Entire. For users who move fast and create repos frequently, this is a significant gap — the most interesting agent sessions (the messy, exploratory POC work) are exactly the ones that never get tracked.

Proposed Behavior

A global configuration mode that automatically enables Entire tracking for any git repo where an agent session starts, without requiring per-repo setup first.

Possible implementations:

• Global agent hooks installed at the user level (e.g., via git config --global core.hooksPath or a Claude Code user-level hook) that detect agent activity and auto-initialize Entire tracking
• entire enable --global flag that sets a user-level default: "track all repos unless explicitly disabled"
• Git template directory integration (init.templateDir) so every new git init or git clone automatically gets Entire hooks
• Agent-side integration — Claude Code (and other agents) could check for Entire and auto-configure on session start if global mode is enabled

Why This Matters

The current per-repo model works well for developers who have a stable set of projects. But for users who:

• Use AI agents as thinking/prototyping tools (not just coding tools)
• Create throwaway repos for POCs, demos, or explorations
• Work across many repos in a single day
• Don't have entire configure as part of their muscle memory

...the friction of per-repo setup means Entire misses the very sessions that would benefit most from being tracked. The "commit log for agents" value proposition is strongest when it's always on, like how shell history works — you don't configure it per-directory.

Context

This came up while testing Entire CLI on a new project created during a Claude Code session. The agent created the repo, built the code, committed, and pushed — but because the session was initiated from a different directory, Entire captured zero checkpoints. The work was done, but the "why" behind it was lost.

Related: #985 (global export/viewing across repos) addresses the read side of this problem. This issue addresses the write side — making sure sessions are captured in the first place.

Environment

• Entire CLI v0.5.x
• Claude Code (claude-opus-4-6)
• macOS

[m13v]

the part that breaks down with per-repo opt-in is the prototyping workflow it's trying to serve never has a 'repo' as the unit. agent sessions span scratch dirs, gh repo create on the fly, and half-finished folders that never get pushed. the actual capture boundary is the agent process, not the working tree. global git template hooks help on git init but miss the case where someone is iterating in /tmp/foo for an hour before deciding it's worth a repo. the write-side fix probably has to live in the agent's session lifecycle, not in git plumbing, otherwise you keep losing the messy exploratory phase that's the whole point.