Fastmail Jmap.Disabled

WarnAudited by ClawScan on May 10, 2026.

Overview

This looks like a real Fastmail email tool, but it asks for broad mailbox/send authority while registry metadata under-declares that access and the package identity/capabilities are inconsistent.

Install only if you intentionally want this agent to have broad Fastmail access. Prefer a dedicated, least-privileged token, verify the publisher/version mismatch is resolved, document or remove the contacts helper, and require confirmation before sending, trashing, or moving email.

Findings (6)

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

If installed with the requested token, the agent can read/search your mailbox, send mail, and change email state such as moving or trashing messages.

Why it was flagged

The skill requires a Fastmail API token with authority to read, modify, and send email. The registry metadata supplied for the skill says required env vars and primary credential are none, so this high-impact permission boundary is under-declared outside the README.

Skill content
Scopes: Email (read/write) + Email Submission (send) ... export FASTMAIL_TOKEN="fmu1-..."
Recommendation

Declare the Fastmail token as a primary credential in metadata, use the least-privileged token possible, and require explicit user approval for send/delete/move operations.

What this means

If the token includes Contacts scope, the agent could retrieve address-book details beyond ordinary email management.

Why it was flagged

The package includes a contacts-access CLI requiring Contacts scope, while the visible top-level description and command list focus on email operations. Contacts can include private phone numbers, addresses, and notes.

Skill content
Fastmail JMAP Contacts CLI — search and read contacts. ... Env: FASTMAIL_TOKEN (API token with Contacts scope)
Recommendation

Either remove the contacts helper or clearly document it in SKILL.md and metadata, including the Contacts scope and what contact fields may be returned.

What this means

It is harder to know exactly which publisher/version you are trusting before giving the skill access to your email account.

Why it was flagged

This conflicts with the submitted registry metadata identifying owner kn750bx66hyza0dw0yxwjat3t181xapp, slug fastmail-jmap-disabled, and version 1.0.0. That mismatch creates a provenance and version-tracking gap.

Skill content
"ownerId": "kn7fqhc36bye71613wpewknkbx81k1mh", "slug": "fastmail-jmap", "version": "1.3.0"
Recommendation

Resolve the registry/_meta identity mismatch and publish from a verifiable source before users provide credentials.

What this means

An agent using these commands can change mailbox state; mistakes could file, hide, mark, or trash messages.

Why it was flagged

The skill documents high-impact email mutation and sending commands. It gives explicit approval guidance for sending, but not for move/trash/mark-read operations.

Skill content
`send <to> <subject> <body>` | Send an email | ... `move <email-id> <mailbox-name>` ... `trash <email-id>` ... Always ask before sending. Never send without approval.
Recommendation

Keep send approval mandatory and add similar confirmation guidance for trashing, moving, or bulk-changing messages.

What this means

A malicious or misleading email could try to influence the agent if the agent treats email text as instructions instead of data.

Why it was flagged

The script returns email body content into the agent/user context. That is expected for an email reader, but mailbox content is untrusted input and may contain instructions aimed at the agent.

Skill content
"fetchTextBodyValues": True, "fetchHTMLBodyValues": True ... print(body_text[:5000] if body_text else "  (no body)")
Recommendation

Treat email bodies as untrusted content; summarize them as data and do not follow instructions embedded in emails unless the user explicitly confirms.

What this means

If enabled in a recurring workflow, the agent may periodically inspect unread mail and surface summaries.

Why it was flagged

The artifact suggests an optional recurring email-check workflow. It does not install persistence itself, but users should understand that configuring this would repeatedly access their mailbox.

Skill content
### In heartbeat/cron: ... Run: `python3 scripts/fastmail.py unread` ... If urgent/actionable emails found, summarize and alert.
Recommendation

Only enable heartbeat/cron checking intentionally, set a clear frequency, and limit what email content may be summarized or alerted.