GitHub Repository

Run approved AI repo changes one item at a time in isolated git worktrees, with verification, dual audits, and explicit human/external stop points.

3 starsPython

Run approved AI repo changes from a Markdown playbook

Name: Run approved AI repo changes from a Markdown playbook
Availability: InStock
Author: EagleEdge

by EagleEdge·Mar 27, 2026·2 points·0 comments

AI Analysis

●●SolidBig BrainShip It

Dual LLM audits on frozen packets with explicit stop states beat fire-and-forget agent loops.

Strengths

•Worktree-per-item isolation means each change attempt is independently auditable and reversible
•Codex and Claude both audit the same frozen packet before merge, catching divergent failures
•Four explicit terminal states (passed, awaiting_human_gate, blocked_external, escalated) enforce control

Weaknesses

•AI code orchestration is increasingly crowded with Cursor, Continue, and Cognition-style tools
•Zero stars and no release tags suggest this is early-stage with unproven adoption