The Agent Skills Directory

[SAFE]: The skill uses Git worktrees to isolate autonomous work from the main codebase, preventing accidental data loss or corruption in the user's primary working environment.
[SAFE]: Prerequisite tools (gh and gw) are standard development utilities. The installation process is transparent and involves user-initiated shell scripts and package managers.
[SAFE]: No evidence of prompt injection, data exfiltration, or obfuscated code was found across the analyzed files. All external links point to official documentation (Anthropic, Google, Claude) or the author's verified repositories.
[SAFE]: The workflow incorporates multiple safety gates, specifically the confidence(plan) gate at Phase 1, which requires a high threshold (90%) and user approval before implementation begins.
[SAFE]: The install.sh script is well-documented and performs standard setup tasks like creating symlinks in the local .claude directory without requiring elevated privileges or executing hidden remote code.

autonomous-workflow