Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent email CLI helper, but users should treat it like any tool that can access, send, and modify email.
Install only if you intend to let Himalaya access your email account. Use a keyring or password manager rather than raw passwords in config, review recipients and attachments before sending, and confirm before deleting, moving, or bulk-changing messages.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
backend.auth.keyring = "imap-example"
Then run himalaya account configure <account> to store the password.
The file includes plaintext password configuration examples in the main minimal setup, which can normalize storing mail credentials directly in config.toml. Users may copy the example into real environments, leaving long-lived IMAP/SMTP credentials exposed to local compromise, backups, logs, or accidental disclosure.
The skill is described as a mail CLI for IMAP/SMTP operations, but this configuration reference documents use of backend.auth.cmd to execute arbitrary shell commands for password retrieval. Spawning local commands is a broader capability than mail handling itself and is not explicitly reflected in the manifest description.
The OAuth2 examples instruct the skill to obtain secrets and tokens by running external commands via *.cmd settings. Executing local commands to fetch sensitive material is a significant capability not justified by the manifest's simple description of email list/read/search/compose/reply/forward/copy/move/delete behavior.
The documentation shows MML attachment and inline-image examples that reference local filesystem paths, but it does not explicitly warn that those paths will be read from disk and included in the outgoing email. In a mail-composition tool, this can lead to accidental exfiltration of sensitive local files if users copy examples or misunderstand what filename=/path/... does.
This file explains how to compose and send messages from the CLI, including template send, but does not include a user-facing warning about the privacy impact of transmitting message bodies, headers, and attachments. For markdown files, omissions of warnings about actions affecting user data or privacy are in scope.
No suspicious patterns detected.