MailMolt - Email for AI Agents
v1.1.0Provides your AI agent with its own email address to send, receive, search, and manage emails independently under human supervision.
⭐ 1· 2k·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
high confidencePurpose & Capability
The handlers, SKILL.md, and hook/config files all implement an email identity service (register, send/receive, search, webhooks) which coheres with the skill's name and intent. However the registry-level metadata shown to the platform lists no required environment variables or primary credential, while the included mailmolt-hook.json, config.json, SKILL.md, and handlers.ts clearly require an API key (mm_live_...) and network access to api.mailmolt.com. That metadata mismatch is an inconsistency that should be clarified.
Instruction Scope
Runtime instructions call only the MailMolt API and instruct storing the agent API key in ~/.config/mailmolt/credentials.json and adding the key to OpenClaw config. They also recommend a 30-minute heartbeat and setting up webhooks to a user-controlled URL. The instructions do not ask the agent to read unrelated system files or other credentials, but recommending writing credentials to a home config file and automatically polling every 30 minutes increases the amount of sensitive data stored and the frequency of external network activity.
Install Mechanism
There is no remote install/download step – the skill is instruction + a local handlers.ts file and manifest entries. No external archives or opaque installers are fetched by the skill itself.
Credentials
The service legitimately requires a secret API key and optionally a webhook secret; that is proportional to an email provider. But the registry metadata omitted required env vars while the included hook.json marks MAILMOLT_API_KEY as required and the SKILL.md instructs storing credentials under ~/.config/mailmolt/credentials.json and in OpenClaw config. The mismatch is suspicious. Also, the agent will receive the human_owner email via heartbeat once available — this is sensitive personal data that MailMolt will expose to the agent and to whatever systems the agent forwards to (webhooks, logs).
Persistence & Privilege
always:false and normal model invocation are used (no forced global residency). The hook.json includes a config suggestion that would add 'mailmolt' to agents.defaults.tools.alsoAllow, which, if applied by an administrator, could make the tool broadly available by default — check whether installing will change global agent defaults. Otherwise the skill does not request elevated OS privileges or modify other skills' configs.
What to consider before installing
What to check before installing:
- Clarify the metadata mismatch: ask the publisher why the registry listing shows no required env vars while mailmolt-hook.json and the code require an API key (MAILMOLT_API_KEY). The platform should list required secrets explicitly.
- Treat the API key as highly sensitive. Prefer storing it in the platform secret store (OpenClaw config secrets) rather than a plaintext file in ~/.config. If you must use a file, restrict filesystem permissions and consider rotating keys.
- Understand what MailMolt will see: the service receives, stores, and returns your agent's messages and (per heartbeat) may return the human owner's verified email. If that data is sensitive, review MailMolt's privacy/retention policy and confirm you trust api.mailmolt.com and the stated repository/homepage (confirm the domain and repo are legitimate).
- Webhooks: if you set up webhooks, provide a verification secret (MAILMOLT_WEBHOOK_SECRET) and validate signatures on your endpoint. Do not expose webhook endpoints that accept unauthenticated requests.
- Human approvals: keep agents in supervised/sandbox mode until you trust behavior. The skill can send email when elevated; limit permission level and require human approval for external sends where possible.
- Audit network usage and logs after enabling: the skill polls heartbeat every 30 minutes and will make API calls for sends/reads/searches; confirm that frequency is acceptable and that logging does not leak message contents.
If you cannot confirm the metadata discrepancy, the origin of the package, or the legitimacy of mailmolt.com/repo, treat the skill as untrusted and do not provide your API key or enable it broadly.Like a lobster shell, security has layers — review code before you run it.
latestvk972bprmmetpw3r5w1vwrjzabh80bqv0
License
MIT-0
Free to use, modify, and redistribute. No attribution required.