Back to skill

Security audit

surface-cli

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed mail-CLI helper whose sensitive mailbox access and write actions are aligned with its stated purpose and guarded by user-directed safety rules.

Install only if you want an agent to work with your email through Surface; review local write-safety settings, prefer drafts or configured test recipients for sends, avoid direct password flags, and confirm any external summarizer before enabling it.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 112)May include surrounding context.

md
`--password <password>` is supported and treats the flag value as the password
directly, but prefer `--password-env`, `--password-file`, or
`--password-command` because direct CLI passwords can leak through shell
history, process listings, terminal logs, or agent transcripts. Do not ask the
user to paste mailbox passwords into chat or store them in the repo.

Local policy lives in:

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 291)May include surrounding context.

md
inbox arrivals plus targeted `search` when the watch has narrow filters.
9. Read only the messages you need with `surface mail read <message_ref>`.
10. For passive watching, do not mutate read state. If the user explicitly asks you to triage unread
   mail and write safety is enabled, marking handled messages read after reporting is acceptable
   unless the user asks to keep them unread.
11. Act using refs from Surface output. Do not rely on array positions from previous JSON.