How SkillRadar scores agent AI skills

SkillRadar is the trust-before-install layer for agent skills: it shows usefulness, risk, provenance, and evidence before a team lets an agent touch files, browsers, credentials, or production workflows.

Current scan snapshot

Report, source, and change context are summarized here instead of living as separate top-level tabs.

128

Skills indexed

Sources tracked

Healthy sources

High-review changes

Latest scan: 2026-06-21 · Change window: 0 total changes, 0 new skills, 0 changed skills.

Maintainer and source reputation

Recent commits, releases, and changelog activity

SKILL.md clarity, examples, and install instructions

Required credentials, env vars, and external API calls

Shell commands, filesystem writes, remote scripts, and dependency surface

Practical utility for developers, operators, and business automation

Immutable evidence

Every high-signal finding should carry the source path, source blob SHA, scanner version, and the exact snippet that triggered review.

Provenance-first scoring

Scores are only useful if teams can trace them back to public source, source blob SHA, generated timestamp, scanner version, and text-only security policy.

Review language, not accusations

Findings are review signals. SkillRadar says what changed and why to inspect it; it does not call projects malware from automated text matches.

Important wording

Risk flags are not accusations. They mean a skill asks for capabilities or includes patterns that deserve human inspection before install, especially in environments with secrets, production data, or messaging access.

Get the signal

Join the SkillRadar launch list.

Daily source refreshes, top skills, risk flags, and early access to private scans, watchlists, and team allowlists.

Prefer the hosted beehiiv page? Open it here →