ClawHub

SWARM Safety

v1.7.1

SWARM: System-Wide Assessment of Risk in Multi-agent systems. 38 agent types, 29 governance levers, 55 scenarios. Study emergent risks, phase transitions, an...

⭐ 0· 412·0 current·1 all-time
by@rsavitt
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description (multi-agent safety research) align with the instructions (pip-installable Python package, simulation API, CLI). The skill does not request unrelated resources (no cloud creds, no unusual binaries).
✓
Instruction Scope
SKILL.md focuses on running local simulations, CLI usage, and a localhost-only API. It explicitly warns not to expose the dev API, not to include real credentials/PII in scenarios, and treats results as research artifacts. The instructions do not direct the agent to read unrelated system files or exfiltrate data.
✓
Install Mechanism
No install spec in registry (instruction-only). SKILL.md recommends standard pip installs or cloning the GitHub repo — common, well-understood install methods with no opaque download URLs.
✓
Credentials
The skill declares no required environment variables, credentials, or config paths. The SKILL.md does show that agent registration returns an api_key for local API use — appropriate and proportional to the stated API functionality.
✓
Persistence & Privilege
always is false and the skill is user-invocable. The SKILL.md says storage is in-memory by default and warns about production deployment changes; it does not request permanent agent-level privileges or modify other skills.
Assessment
This skill appears coherent with its stated purpose, but standard caution is advised: (1) confirm the PyPI package name and the GitHub repo (https://github.com/swarm-ai-safety/swarm) match what you expect before installing; (2) install in a virtual environment or sandbox; (3) do not bind the API to 0.0.0.0 or expose it to untrusted networks; (4) do not include real API keys, passwords, or PII in scenario YAMLs; (5) if you need higher assurance, review the package source on GitHub or the PyPI release contents and checksums before installing.

Like a lobster shell, security has layers — review code before you run it.

latestvk972cbxmec0xsw9n1fqa70frqn81q94w

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments