@symbo.ls/cli@3.14.817
contains an exposed credential; upgrade to >=3.14.839
Fetch your Symbols configuration
AI security review
scanned · by lpm-firewall-aiThe source confirms that one-shot Claude requests can execute model-requested shell commands without per-command confirmation, which warrants a warning for dangerous agent capability. The manifest has no install hook, and the inspected login and dependency excerpts do not establish credential theft or account manipulation.