The default MCP initialization silently installs hooks and rules into existing AI-agent configuration directories, changing behavior beyond the server’s own process. The hook data flow matches the documented memory feature and sends the SlashML key to its own service, so this warrants a warning rather than a malicious verdict.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
An MCP client initializes the server, unless auto-hooks are disabled by environment setting.
Impact
Agent behavior is changed across future sessions, and hook event data is sent to the cloud service as part of memory assimilation.
Mechanism
The server silently installs hooks and rules in detected agent configuration directories. Later hook events send prompts and tool-event data to the SlashML memory API.
Evidence
bin/slashml-mcp.jsREADME.mdpackage.json~/.claude/settings.json~/.claude/CLAUDE.md~/.cursor/hooks.json~/.cursor/rules/slashml-memory.mdc~/.codeium/windsurf/hooks.json~/.codex/hooks.json~/.codex/AGENTS.md