# Agentic Security Field Guide

> Find the right AI security tool for the job: curated AI agents, skills and MCP servers, each described with what it does, what it needs and what its code can touch. Task pages compare the tools for one task side by side. Rights are read three ways: what the author declares, what the code can do, and, once a tool has run in the lab, what it did.

## Method and data

- [Method](https://readbeforerun.ai/method.md): what is checked and how to use the result
- [Rules of the method](https://readbeforerun.ai/method/rules.md): selection criteria, Privilege Reach scale, lab readings, evidence rules
- [catalog.json](https://readbeforerun.ai/catalog.json): all entries with reach as text, CC BY 4.0
- [Entry schema](https://readbeforerun.ai/schema/entry.schema.json), [observation schema](https://readbeforerun.ai/schema/observation.schema.json), [catalog schema](https://readbeforerun.ai/schema/catalog.schema.json)
- [Feed](https://readbeforerun.ai/feed.xml), [JSON Feed](https://readbeforerun.ai/feed.json), [field notes](https://readbeforerun.ai/fieldnotes.xml), [catalog changes](https://readbeforerun.ai/catalog-changes.xml)

## Recently tried

- [Cisco MCP Scanner Plugin](https://readbeforerun.ai/catalog/cisco-mcp-scanner.md): reach 3 · beyond declared (1); code shows: reads credential stores · talks to any host · runs shell commands or eval · +2 more; Beyond declared on 1 of 6 rights: network (observed 2, declared 1), with blocked attempts to reach GitHub at start-up even in YARA-only mode.
- [CrowdStrike Falcon MCP Server](https://readbeforerun.ai/catalog/crowdstrike-falcon-mcp.md): reach 3 · beyond declared (1); code shows: can tunnel or exfiltrate · writes outside the project · changes config or installs globally · +1 more; Beyond declared on 1 of 6 rights: secrets (observed 3, declared 1), reading ~/.netrc on every API request.
- [GitGuardian MCP Server](https://readbeforerun.ai/catalog/gitguardian-ggmcp.md): reach 3 · as declared; code shows: can tunnel or exfiltrate · writes outside the project · reads broad tokens or the environment · +2 more; Within declared on every right the lab could observe.

## Optional

- [Catalog page](https://readbeforerun.ai/catalog/)
- [All entries as Markdown](https://readbeforerun.ai/catalog/): append .md to an entry path, for example https://readbeforerun.ai/catalog/<slug>.md