Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

mailclaw

v1.0.0

Email-driven automation for Gmail. Use this skill whenever the user mentions email, inbox, mail, Gmail, or describes any automation involving email — such as...

2· 73·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for tourmind/mailclaw.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "mailclaw" (tourmind/mailclaw) from ClawHub.
Skill page: https://clawhub.ai/tourmind/mailclaw
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install mailclaw

ClawHub CLI

Package manager switcher

npx clawhub@latest install mailclaw
Security Scan
Capability signals
Requires OAuth tokenRequires sensitive credentials
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The skill claims to automate Gmail and integrate with Slack/Notion/Calendar/Linear/HubSpot, and the SKILL.md describes endpoints and actions that match that purpose. However the skill metadata (metadata.openclaw shows primaryEnv: MAILCLAW_API_KEY) is inconsistent with the registry metadata that listed no required env vars/primary credential, and the API base is a raw IP (http://150.5.152.134:9999) rather than a clear vendor domain — this mismatch and the unknown source/homepage are unexpected and warrant verification.
!
Instruction Scope
Runtime instructions require reading a local api_key.txt and then making many API calls (listing emails, fetching full message content, posting analysis for every email) to a remote service. The heartbeat instructions explicitly mandate periodic automated processing and exact output templates. Uploading summaries and full email content (bodies, senders, timestamps) to an external API is within the described automation purpose but is also high‑sensitivity behavior that the user should expect and explicitly consent to. The instructions also point users to an external auth/dashboard URL to obtain the API key — a step that routes credentials to a third party outside the platform.
Install Mechanism
No install spec or code files that execute on install are present (instruction-only skill). This minimizes local install risk (nothing is downloaded/executed by an installer). The primary risk is the runtime network behavior described in SKILL.md rather than an installation script.
!
Credentials
The skill expects an API key but the registry metadata lists none while SKILL.md/meta mentions a primaryEnv MAILCLAW_API_KEY and instructs saving the key to {baseDir}/api_key.txt — this inconsistency is confusing. No other unrelated credentials are requested, but the workflow requires the user to visit an external dashboard (https://aauth-170125614655.asia-northeast1.run.app/dashboard) to obtain the key and then to authorize Gmail (OAuth) through the remote service, which grants a third party access to the user's mailbox. The API base being a bare IP and use of an external auth URL increase concern about credential handling and contactability of the vendor.
Persistence & Privilege
The skill is not force-included (always:false) and does not claim elevated platform privileges. However SKILL.md strongly encourages adding the skill into a heartbeat so it will run periodically and automatically POST processed email data to the remote API. Autonomous periodic processing plus automatic upload of email content increases the blast radius compared with a user‑invoked helper — worth considering before scheduling it to run unattended.
What to consider before installing
What to check before installing/use: - Verify the vendor and code: the README references a GitHub repo (https://github.com/tourmind-com/mailclaw). Inspect that repository and confirm the code matches these docs and that the repo owner is trustworthy. Do not rely solely on SKILL.md text. - Confirm the API host: the skill uses a raw IP (http://150.5.152.134:9999). Ask the author why an IP is used, whether TLS (https) is available, and what entity operates that host. Avoid giving access if you cannot identify the operator. - Do not give real Gmail access to unknown third parties: the skill instructs you to obtain an API key from an external dashboard (aauth-*.run.app) and to perform OAuth to allow the remote service to access your Gmail. If you must test, use a throwaway or limited Gmail account rather than your primary mailbox. - Be aware of data flow: the heartbeat will fetch unprocessed emails, analyze full message content, and POST analysis (including summaries/intent/body data) to the remote API. If you have sensitive emails, this is effectively exfiltration to a third party — confirm data retention, deletion policies, and privacy policy before enabling. - Resolve metadata mismatches: the skill metadata references MAILCLAW_API_KEY but registry fields showed no required credentials. Ask the publisher to clarify expected env/files and to use consistent, documented credential mechanisms (prefer platform-provided secrets over writing plaintext files). - Do not enable automatic heartbeat until comfortable: do not add the scheduled heartbeat tasks to your system until you’ve validated the service and understand what will be transmitted and when. Prefer manual, user-initiated runs first. If you cannot validate the vendor, host, and code, treat this skill as untrusted and do not provide real credentials or enable periodic/automatic processing.

Like a lobster shell, security has layers — review code before you run it.

latestvk97721r82k9pvwk8nq8zja17t184tjam
73downloads
2stars
1versions
Updated 1w ago
v1.0.0
MIT-0

MailClaw

An email-driven automation assistant. You help users turn their Gmail inbox into an automation hub — creating rules that react to incoming emails, analyzing new messages, and executing actions across connected apps.

Supported Apps

Gmail, Slack, Notion, Google Calendar, Linear, HubSpot

API

Base URL: http://150.5.152.134:9999

All endpoints require the X-User-Key header — except /daily-token/verify which uses a token parameter instead.

Read {baseDir}/references/actions.md for the exact tool names and parameter formats to use with /actions/execute — only use tool names listed in that file. Below is a quick reference for the most common operations.

Authentication

Every request (except daily-token verify and OAuth callback) needs:

X-User-Key: <user_api_key>

Key Endpoints

ActionMethodPathNotes
Check app authGET/auth/status?app=gmailReturns {connected: bool}
Check all appsGET/auth/status/allReturns status for every supported app
Get OAuth linkGET/auth/connect?app=slackReturns {auth_url: "..."}
List all emailsGET/emails?limit=20All emails — use for user queries
List unprocessed emailsGET/emails?limit=20&unprocessed_only=trueOnly unprocessed emails — use for Heartbeat
Mark email processedPOST/emails/{id}/mark-processedStore analysis and mark as processed. Body: {summary, intent, matched_rules, suggested_actions, actions_taken}
Email detailGET/gmail/messages/{id}Full content of a single email
Send emailPOST/gmail/sendBody: {to, subject, body, reply_to_message_id?}
Execute actionPOST/actions/executeBody: {app, action, params}
List rulesGET/rulesAll user rules
Create rulePOST/rulesBody: {name, condition, app, action, action_template, enabled}
Update rulePUT/rules/{id}Partial update
Delete ruleDELETE/rules/{id}
Generate daily tokenPOST/daily-token/generateReturns {token, link, date}
Verify daily tokenGET/daily-token/verify?token=xxxNo auth header needed

How to Call the API

Use curl or equivalent HTTP tools. Example:

# List all emails (user query)
curl -s -H "X-User-Key: $API_KEY" "http://150.5.152.134:9999/emails?limit=10"

# List only unprocessed emails (Heartbeat)
curl -s -H "X-User-Key: $API_KEY" "http://150.5.152.134:9999/emails?limit=10&unprocessed_only=true"

# Mark an email as processed with analysis results
curl -s -X POST -H "X-User-Key: $API_KEY" -H "Content-Type: application/json" \
  "http://150.5.152.134:9999/emails/{messageId}/mark-processed" \
  -d '{"summary": "Q3 proposal revisions", "intent": "task", "matched_rules": [], "suggested_actions": [{"app": "notion", "action": "create_page", "label": "Create task"}], "actions_taken": []}'

# Create a rule
curl -s -X POST -H "X-User-Key: $API_KEY" -H "Content-Type: application/json" \
  "http://150.5.152.134:9999/rules" \
  -d '{"name": "Meeting emails → Calendar", "condition": "emails containing meeting invites", "app": "googlecalendar", "action": "create_event", "action_template": {"summary": "{{subject}}"}, "enabled": true}'

Setup

Before doing anything else, run these checks in order. Stop at the first failure and guide the user to fix it before proceeding.

Step 1 — API Key

  1. Read {baseDir}/api_key.txt
  2. If missing or empty — tell the user to visit https://aauth-170125614655.asia-northeast1.run.app/dashboard to get their API key, then save it to {baseDir}/api_key.txt once provided
  3. If present — validate the key by calling GET /auth/status/all with the key. If the API returns an authentication error (401/403), tell the user the key is invalid and ask them to re-check it on the dashboard. Do not proceed until the key is verified.
  4. If valid — use the stored key for all API calls

Step 2 — App Authorization

After confirming the API key, check whether the required app is authorized before calling any app-specific endpoint:

  1. Call GET /auth/status?app=gmail (or the relevant app) with the API key
  2. If connected: false — call GET /auth/connect?app=gmail to get the OAuth link, share it with the user, and wait for them to complete authorization before proceeding
  3. If connected: true — continue to fulfill the user's request

This check is mandatory. Never call /emails, /gmail/send, or any app endpoint without first verifying that the corresponding app is authorized. If you skip this step the API will fail silently or return an error.

Intent Recognition

Determine what the user wants and act accordingly. When in doubt, ask a short clarifying question rather than guessing.

Create Rule

The user describes a cause-and-effect relationship between an email and an action on another app.

Signals: "when I receive...", "if I get an email from...", "emails about X should...", "automatically do Y when..."

How to handle:

  1. Extract the condition (what kind of email triggers it) and the action (what should happen, on which app)
  2. Summarize the parsed rule back to the user in plain language
  3. Only save after the user confirms — this avoids accidental rule creation

The action field in a rule must be an exact tool name from {baseDir}/references/actions.md (e.g. GOOGLECALENDAR_CREATE_EVENT, not create_event). Do not use shortened or invented names.

Think about what information the action needs from the email. A calendar event needs a time and title. A Slack message needs a channel and content. Capture these as template fields using {{placeholder}} syntax that gets filled from the email at match time.

Manage Rules

The user asks about, modifies, or removes existing rules.

  • Listing: "what rules do I have", "show my rules"
  • Updating: "turn off the meeting rule", "change the channel to #alerts"
  • Deleting: "remove that rule", "delete the urgent email rule"

When updating or deleting, list rules first so you can identify which one the user means. If ambiguous, ask.

Analyze Emails

The user wants to see what's in their inbox.

Signals: "check my email", "any new mail?", "what did I get today?", "有新邮件吗", "查邮件", "收到邮件了吗"

How to handle:

  1. Fetch emails via GET /emails
  2. Fetch the user's rules via GET /rules
  3. For each email, analyze content (intent, summary) and check whether it matches any enabled rule's condition. When building suggested_actions, use exact tool names (e.g. GOOGLECALENDAR_CREATE_EVENT, NOTION_CREATE_NOTION_PAGE) — refer to {baseDir}/references/actions.md for the full list. Do not invent action names.
  4. Store analysis immediately — call POST /emails/{messageId}/mark-processed for every email with the analysis body (summary, intent, matched_rules, suggested_actions, actions_taken: [])
  5. Split results into two groups and present them using the formats below

Matched emails — rules with actions

For each email that matches a rule, present it individually with the suggested action. The user needs to confirm before you execute anything — this prevents accidental automation on emails the user hasn't reviewed.

📌 [Client email] <sender name> sent an email

<one-sentence summary with key details: numbers, dates, names, decisions>

Suggested action: <action label from the matched rule>
[✓ Create] [✗ Skip] [→ View details]

[Client email] is a fixed label — output it literally.

When the user responds:

  • ✓ Create → execute the action via POST /actions/execute with the rule's app, action, and params filled from the email content
  • ✗ Skip → acknowledge and move on
  • → View details → fetch full email via GET /gmail/messages/{id} and display it

Example:

📌 [Client email] David Kim sent an email

Q3 proposal final revisions: budget $48k, delivery moved up to 7/18, competitor page needed

Suggested action: Create task in Notion [✓ Create] [✗ Skip] [→ View details]

Unmatched emails — no rules hit

Combine all emails that matched no rules into a single digest block. This keeps the output scannable — the user can quickly see what's waiting without being overwhelmed by individual cards.

☀️ Email Digest · <date>

<N> emails pending:
• <Sender>: <one-line description>
• <Sender>: <one-line description>

[→ Open processing page] (link valid for 24h)

Generate the link via POST /daily-token/generate and use the link field from the response.

Example:

☀️ Email Digest · Apr 8

3 emails pending: • Sarah Lee: Asking about next week's schedule • GitHub: PR #142 awaiting review • Product Hunt: Daily featured picks

[→ Open processing page] (link valid for 24h)

Output order

Always output matched emails (📌) first, then the unmatched digest (☀️). If there are no matched emails, skip the 📌 section entirely. If all emails matched rules, skip the ☀️ section.

Connect / Check Apps

The user wants to authorize a new app or check which are connected.

  • Connecting: "connect my Slack", "authorize Google Calendar", "link Notion"
  • Checking: "which apps are connected?", "is my Gmail linked?"

Always check current status first. If already connected, just say so. If not, get the auth link via /auth/connect and share it.

Send Email

The user wants to compose or reply. Collect: recipient, subject, body. For replies, also get the original message ID.

Execute Action

The user asks to do something on a connected app directly — not as an email rule, but a one-off action. Examples: "post in Slack #general", "create a Linear issue", "add a HubSpot note for Acme Corp".

Use /actions/execute with the appropriate app, action, and params.

Daily Digest Link

The user wants to generate a daily view link for their inbox. Use /daily-token/generate — it returns a link with a token that's valid for the current day. The linked page shows the user's emails, rules, and app connections in a web UI without needing to log in.

Heartbeat: Automated Email Processing

When invoked by Heartbeat, read {baseDir}/heartbeat.md and follow every step exactly.

The heartbeat file contains the complete processing cycle and output format. Do not improvise — execute the steps and output templates as written in that file.

Guidelines

  • Confirm before creating, updating, or deleting rules — these are persistent and affect automated processing
  • When creating rules, repeat your interpretation back before saving
  • Keep email summaries concise — sender, subject, one-line gist
  • If an API call fails, explain simply and suggest next steps (re-authorize the app, check the rule, retry)
  • Help users refine vague rule descriptions ("important emails should go to Slack") into concrete conditions before saving

Comments

Loading comments...