swarm-ai-researchOfficial·16 Agent Skills Included

swarm

Simulate multi-agent systems and measure emergent safety risks

Simulates multi-agent ecosystems to measure emergent risks like deception, collusion, and adverse selection using probabilistic safety metrics. Runs governance experiments, parameter sweeps, and red-team scenarios with reproducible seeds and statistical rigor. Eliminates manual experiment tracking with automated run indexing, claim verification, and publication-ready plots and paper scaffolding.

39|4

↗

npx skills add swarm-ai-research/swarm --all -g -y

Available:

Directs the agent to run simulations, sweeps, and research workflows using the repo's slash commands, specialist roles, and memory system while keeping results reproducible.

All Skills in This Repository (16)

Pure Emerald Level Indicators

📦 In Repo

swarm-ai-research

swarm

Measure emergent failures in multi-agent systems using Python.

statistical-analysis

Analyze SWARM experimental data with hypothesis tests and multiple-comparison corrections.

plotting

Create bar charts, box plots, and time-series plots from SWARM simulation data.

parameter-sweep

Automate parameter grid sweeps across SWARM safety scenarios and generate summary statistics.

run-scenario

Execute predefined SWARM simulation scenarios and export standardized results.

paper-writing

Generate markdown research paper skeletons from SWARM experiment data.

regression-check

Compare AI agent simulation metrics against a baseline to identify deviations.

verify

Run integrity checks on vault schema, evidence, links, index, and claims.

session-close

Summarize research work, update memory logs, commit changes, and push to version control.

run-query

Query the run index and vault for experiment history by tags, dates, types, or claims.

kb-query

Query a structured SWARM knowledge graph for related pages, backlinks, and paths.

vault-init

Initialize and extend SWARM Research OS vaults with directories and schema templates.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install SWARM?▼

Run `npx skills add swarm-ai-research/swarm --all -g -y` in your terminal to install all skills in this suite globally.

What does SWARM measure?▼

It measures emergent risks in multi-agent systems, such as toxicity, quality gaps, collusion, and incoherence, using soft probabilistic labels instead of binary good/bad classifications.

How do I run a simulation scenario?▼

Use the run-scenario skill or the CLI command `swarm run scenarios/baseline.yaml --seed 42` to execute a scenario and export standardized metrics and artifacts.

Can SWARM test governance mechanisms?▼

Yes. It supports governance levers like transaction taxes, circuit breakers, audits, staking, and collusion detection, with parameter sweeps to compare their safety and welfare trade-offs.

Does SWARM support statistical analysis of results?▼

Yes. Built-in skills run Welch's t-tests, Cohen's d effect sizes, and Bonferroni corrections, and generate publication-quality plots and paper drafts from experiment data.