swarm-orchestrator
VerifiedAudits AI-generated pull requests to catch shortcuts and incomplete code changes.
What is swarm-orchestrator?
Swarm Orchestrator is an open-source CLI agent that analyzes code changes from AI coding tools. It identifies eleven categories of shortcuts that make patches appear complete without actually solving the underlying problem.
The system runs a multi-persona pipeline against typed obligation contracts and offers eight proof protocols for execution-grounded verification. Results can be advisory or configured to block merges when self-certifying runtime evidence is present.
It is designed for engineers reviewing high volumes of AI-written PRs, compliance teams needing traceable AI-BOM outputs, and maintainers who require hash-chained audit records for every graded patch.
What you can build with swarm-orchestrator
PR Quality Gate
Run in gate mode to block merges unless execution proofs confirm that all detected cheats have been resolved.
Compliance Reporting
Generate CycloneDX-ML or SPDX 3.0 AI-BOM artifacts that map directly to regulatory annexes for EU AI Act and CISA submissions.
Local Diff Review
Pipe any local diff through the full detector set to surface issues before opening a pull request.
Install swarm-orchestrator
git clone https://github.com/moonrunnerkc/swarm-orchestrator.git && cd swarm-orchestrator && npm install && npm run build && npm linkgit clone https://github.com/moonrunnerkc/swarm-orchestrator.git
cd swarm-orchestrator
npm install
npm run build
npm link
swarm --help- 1Clone the repository and install dependencies with npm.
- 2Build the project and link the binary globally.
- 3Set a GITHUB_TOKEN and run swarm audit on a PR reference.
- 4Review flagged issues in advisory mode or enable gate mode for blocking behavior.
- 5Optionally emit an AI-BOM artifact or run in shadow mode for silent logging.
swarm-orchestrator: pros & cons
Pros
- +Catches structural cheats that standard linters miss
- +Provides reproducible runtime proofs instead of just pattern matches
- +Generates regulatory-grade AI-BOM artifacts with minimal setup
- +Supports both lightweight advisory and strict enforcement workflows
Cons
- –Requires Node 20+ and manual configuration for gate mode
- –Proof protocols are currently limited to eight specific cheat types
- –GitHub token and repository access are needed for full PR integration
Frequently asked questions
Only in gate mode when a self-certifying runtime proof fails; the default advisory mode never blocks.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other general-purpose options worth comparing.
