agent-capmap
See what your AI agents can actually do, and which combinations of their MCP tools an attacker could abuse.
1 · Read
Private data
files, repos, email, databases
+
2 · Ingest
Untrusted content
web pages, issues, inboxes
+
3 · Send
Data out
email, HTTP, git push
= the lethal trifecta: injected text can leak your data
MCP
policy as code
SARIF
pipx install agent-capmap  ·  github.com/NKwolv/agent-capmap