reversa-autonomous

Runs the full Reversa agent sequence end-to-end without intermediate confirmations.

1.5k|405|Updated Apr 26, 2026
One-click install
npx skills add https://github.com/sandeco/reversa --skill reversa-autonomous
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: reversa-autonomous
Source: https://github.com/sandeco/reversa/tree/main/agents/reversa-autonomous
Command: npx skills add https://github.com/sandeco/reversa --skill reversa-autonomous

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Running the Reversa reverse-documentation workflow normally requires answering questions and confirming each agent handoff, which blocks unattended sessions. This Skill collects every decision in a single upfront interview and then executes the entire agent pipeline autonomously, ideal for YOLO-mode or unsupervised sessions.

Core Features & Use Cases

  • Single upfront interview: Gathers installation data, documentation level, spec organization, and gap-handling preferences in one interactive session before execution begins.
  • Autonomous orchestration: Inherits the full Reversa orchestrator plan (Scout, per-module agents, regression check) and runs it sequentially with automatic handoffs and mandatory state checkpoints.
  • Closed list of legitimate stops: Only pauses for migration decisions, user-requested chat questions, unrecoverable errors, non-destructive rule risks, or context overflow.
  • Use Case: Point it at a legacy codebase in a Claude Code YOLO session, answer one interview, and return later to find complete traceable specifications in the output folder plus a final report with confidence counts and open questions.

Quick Start

Run /reversa-autonomous in the root of the legacy project and answer the initial interview to start the full unattended specification run.

Frequently Asked Questions about reversa-autonomous

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run Reversa without stopping between agents?▼

Invoke /reversa-autonomous instead of /reversa. It collects all decisions in one initial interview, then runs the full agent sequence end-to-end with automatic handoffs and no intermediate confirmations.

What is the difference between reversa and reversa-autonomous?▼

reversa-autonomous inherits the same plan and agent sequence as the reversa orchestrator but front-loads all questions into a single interview and overrides confirmation prompts, checkpoint offers, and conflict menus with safe defaults.

Does reversa-autonomous work with Claude Code YOLO mode?▼

Yes, it is designed for sessions with automatic tool approval such as Claude Code YOLO mode, and also supports Codex, Cursor, and Gemini CLI. It enforces a strict non-destructive rule since no human approves each action.

When does reversa-autonomous pause execution?▼

It only stops for a closed list of cases: pending migration decisions, user-chosen chat question mode, unrecoverable errors like corrupted state.json, non-destructive rule risks, or context overflow after saving a checkpoint.

Can I resume an interrupted reversa-autonomous run?▼

Yes, checkpoints are saved to .reversa/state.json after every agent. On resume it re-asks only unanswered interview questions, shows progress, and continues the next pending plan task without requesting confirmation.