The flagged network and credential flows belong to the user-invoked AI command, which sends keys to their corresponding provider APIs. The only lifecycle script is prepublishOnly and builds the package; the reviewed source showed no malicious automatic behavior.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
A user runs a command such as ask, login, or tunnel; AI requests occur after the user invokes ask.
Impact
No automatic credential exfiltration, unrelated account action, or install-time data collection was found.
Mechanism
The ask command sends prompts to the selected AI provider and uses that provider's configured key with its corresponding API. The package also supports user-configured Symbols API and tunnel connections.
Evidence
~/.smblsrc~/.claude/claude_desktop_config.json~/.cursor/mcp.json~/.windsurf/mcp.json.symbols_local/chat/*.json
Network endpoints6
api.anthropic.com/v1/messagesapi.openai.com/v1/chat/completions