SR
SkillRadar
Agent security + benchmarks
Menu⌄

Security first

Safety Radar beta now; performance benchmarks, private registries, and runtime skill recommendations next.

Free concierge beta

Submit your agent stack for a free Safety Radar scan.

Paste your Claude Code, Cursor, MCP, OpenClaw, plugin, or skill URLs. SkillRadar watches the instruction layer for prompt injection risk, new shell commands, credential requirements, external domains, installers, and risk-score changes.

Get my free Safety Radar scanSee sample alert

Safety Radar

Prompt injection risk, hidden installers, shell commands, credential requirements, and risky diffs.

Performance Radar

Planned benchmarks for task success rate, token efficiency, speed, and model compatibility.

Private Radar

Enterprise path for approved internal skill registries, policies, and team governance.

AgentSkillOS

Runtime skill-stack recommendations so agents choose safer, better-fit tools without instruction bloat.

What happens after you submit

Built for the first 10 beta teams.

  1. 1Submit your agent stack: Claude Code, Cursor, MCP, OpenClaw, plugins, skills, or GitHub URLs.
  2. 2SkillRadar turns it into a private Safety Radar watchlist and stores the request in the Google Sheet intake queue.
  3. 3You get a concise risk/change report first; benchmarking and automated delivery come after demand is validated.
Early version: Safety Radar reports are generated from sheet-backed watchlists first. Performance Radar benchmarks, automated email delivery, private registries, and runtime APIs come after demand is validated.

Safety Radar beta intake

Submit your agent stack.

Paste your Claude Code, Cursor, MCP, OpenClaw, plugin, or skill URLs. In exchange for your email and stack URLs, we set up a free Safety Radar scan/watchlist and send the first risk/change report.

Alert me when these change

Safety posture

Text-only alerts, not code execution.

SkillRadar treats third-party skills as hostile text: no code execution, no dependency installs, no loading scanned skills into an agent runtime, and no secrets in the scanner environment. Alerts are review prompts for humans before install or automation.