Himalaya

PassAudited by ClawScan on May 10, 2026.

Overview

This is a coherent email-CLI skill, but it can access and change a configured mailbox, so sensitive actions should stay user-approved.

Install this skill only if you want the agent to use Himalaya for your email. Prefer app passwords, OAuth, `pass`, or keyring storage; avoid raw passwords; verify which account is active; and require explicit approval before sending, replying-all, forwarding, deleting, moving, or downloading attachments.

Findings (5)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A configured agent session could read email and send messages as the configured account.

Why it was flagged

The skill requires credentials for a real email account so the CLI can authenticate and act on the user's mailbox. This is expected for the stated purpose, but it is privileged account access.

Skill content
IMAP/SMTP credentials configured (password stored securely)
Recommendation

Use app passwords, OAuth, `pass`, or a system keyring where possible; avoid raw stored passwords; and only configure accounts you are comfortable letting the agent access.

What this means

Accidental or over-broad use could send unwanted emails, reply-all, delete messages, move messages, or change flags.

Why it was flagged

The documented CLI operations can send email and mutate mailbox state. These commands are purpose-aligned examples, but they are high-impact if run without user intent.

Skill content
`himalaya template send` ... `himalaya message delete 42` ... `himalaya message move 42 "Archive"`
Recommendation

Require explicit confirmation before sending, forwarding, replying-all, deleting, moving, or bulk-changing messages, and verify the target account, folder, and message IDs first.

What this means

Emails and attachments may contain sensitive data or prompt-injection text that tries to influence the agent.

Why it was flagged

Reading, exporting, and downloading email brings private and untrusted message content into the agent's working context or local files.

Skill content
`himalaya message read 42` ... `himalaya message export 42 --full` ... `himalaya attachment download 42 --dir ~/Downloads`
Recommendation

Treat message bodies and attachments as untrusted data, do not follow instructions found inside emails, and only download or expose messages the user requested.

What this means

A malicious or mistaken config change could cause the CLI to run an unintended local command.

Why it was flagged

Himalaya can run a local command to retrieve a password. The examples use standard password-manager commands, but this config field is still a local command-execution surface if modified by untrusted input.

Skill content
`backend.auth.cmd = "pass show email/imap"`
Recommendation

Keep `auth.cmd` values limited to trusted password-manager commands and protect the Himalaya config file from untrusted edits.

What this means

The installed `himalaya` binary will handle mailbox credentials and email operations.

Why it was flagged

The skill delegates behavior to an external Homebrew-installed binary rather than code included in the skill artifact. This is normal for a CLI skill but depends on trusting that package source.

Skill content
brew | formula: himalaya | creates binaries: himalaya
Recommendation

Install from trusted package sources and keep the CLI updated.