skills/qwibitai/nanoclaw/add-gmail/Gen Agent Trust Hub

add-gmail

Warn

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill adds a remote git repository from the vendor (https://github.com/qwibitai/nanoclaw-gmail.git) and fetches/merges code into the local project environment.- [REMOTE_CODE_EXECUTION]: Uses npx to download and execute the package @gongrzhe/server-gmail-autoauth-mcp. This package originates from an unverified third-party user on the NPM registry and is used to handle sensitive OAuth authentication flows.- [COMMAND_EXECUTION]: Orchestrates several shell commands including git fetch, git merge, npm install, and service restarts via launchctl or systemctl to apply changes.- [DATA_EXFILTRATION]: Reads and writes sensitive Gmail API credentials and OAuth tokens to the ~/.gmail-mcp/ directory. While necessary for functionality, this directory is mounted into the agent container, creating a sensitive data surface.- [PROMPT_INJECTION]: Implements a capability to poll and process external email data, which acts as an indirect prompt injection vector.
  • Ingestion points: Unread emails from the Gmail inbox (monitored in Phase 4).
  • Boundary markers: Present; the skill appends instructions to CLAUDE.md advising the agent not to reply to emails automatically.
  • Capability inventory: Tools for reading, sending, searching, and drafting emails (Phase 1).
  • Sanitization: Absent; the skill does not specify any sanitization or filtering of email body content before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 12, 2026, 02:34 PM