ADR vs geiger
Uber's Agentic AI Detection and Response: inventory AI tools on endpoints, capture Claude Code/Cursor/Codex telemetry, and flag risky sessions with a two-tier detector; ships ADR-Bench (300+ tasks). — versus — Read-only inventory of every AI agent, MCP server, hook, plugin and AI extension on a machine: npx geiger-scan labels what can execute code, hold secrets or reach the filesystem, with evidence.
Both inventory the AI agents, IDE extensions, local runtimes and MCP servers present on a machine and flag what needs review. geiger is a one-shot npx scan; ADR Discovery feeds the same inventory into a fleet-wide detection pipeline.
| ADR | geiger | |
|---|---|---|
| Stars | 2.1k | 171 |
| Forks | 227 | 9 |
| Language | Python | JavaScript |
| License | Apache-2.0 | MIT |
| Last activity | 6 days ago | 1 months ago |
| Topics | security, evals | security |
| Curated connections | 3 | 7 |
ADR — the curator's take
An enterprise SOC's starting kit for coding agents, with production credibility (deployed at Uber, MLSys 2026). Take the pieces you need: Discovery for shadow-AI inventory, Sensor for a unified telemetry schema across 7+ harnesses, Detector for cheap triage plus agentic deep review, ADR-Bench (134 MCP servers, 17 attack techniques) to score any detector you are considering. Do NOT expect blocking — Prevention is unreleased, so it tells you what happened, not stops it; pair with a runtime guard. Also skip if you are one developer: the default detector needs Anthropic and OpenAI keys and the value shows at fleet scale.
geiger — the curator's take
Run it once on any laptop that has had Claude Code, Cursor, Codex, Goose or a dozen MCP servers installed and you will find things you forgot about. Zero deps, no telemetry, never prints secret values (shape only), `--strict` exits non-zero for CI or fleet checks, `--home` scans another profile or a mounted image. It is a static inventory — it tells you what is *configured* and what it could reach, not what it *did*; pair it with a runtime layer (adrian) for that. Not a policy enforcer and not a skill scanner (asm does that). Tiny project (~100 stars) but the detector table is broad and the promises are the right ones.