RedMuse

RedMuse - Autonomous pentesting, built for Muse.ai.

Autonomous pentesting, built for Muse.ai, Meta's personal AI agent.
No Docker, no LLM key, no separate engine. Muse is the engine: an
orchestrator plus specialist agents run recon, exploitation,
proof-of-concept validation, fixing, and reporting against targets
you own or are authorized to test.

Inspired by the architecture of Strix
(agent graph, skill packs, sandbox runtime, SARIF-style reporting),
rebuilt from scratch as skills and playbooks that anyone with Muse.ai can
install and run. Skill content here is original writing. Strix itself
is Apache-2.0, and its methodology topics (OWASP Top 10, API Top 10,
IDOR, SSRF, and friends) are credited in docs/ARCHITECTURE.md.

Quick start

In Muse.ai: no terminal needed

You do not clone anything. Ask Muse to install the pack, and it
fetches the skills from the repo. Paste this, replacing the
target with your own:

Install the RedMuse skill pack from https://github.com/reblox01/RedMuse
and run a quick scan of https://staging.example.com.
I own this target and authorize active testing on it.
Start with the authorization gate, then run the phases and write
the report.

Same install, other jobs:

Install the RedMuse skill pack from https://github.com/reblox01/RedMuse
and run a dev-audit of my code against my localhost staging copy.
Code audit first, then a live proof for each finding, then fix
and retest.

Install the RedMuse skill pack from https://github.com/reblox01/RedMuse
and fix the open findings in redmuse-runs/<run-name>/, one finding
at a time, retesting each original proof before closing it.

No active test runs without the gate: the orchestrator records
target, scope, and authorization first.

On your own PC: terminal install

Only for local setups with a terminal and a workspace folder,
such as a l