Observed arrival · 2026-11-21
AgentLens scans the instructions agents trust
A read-only scanner for AI agent files and repositories, looking for prompt injection, exposed credentials, and unsafe commands.
- For
- People inspecting AI agent instructions and skills
- Worth noticing
- High- or critical-risk snapshots can be registered indefinitely, including scans of private repositories and pasted content.
Field notes
The scanner supports individual files, pasted SKILL.md content, folders, and GitHub repositories, with folder uploads preserving paths such as .claude/settings.json. Repository scans are limited to supported agent files rather than the whole repository; GitHub directory URLs can further restrict the scan. The page also distinguishes complete supported skill versions from partial MCP tool-definition snapshots, which cover saved definitions rather than running servers.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue