Back to skill

Security audit

SRIFT Peerknock

Security checks for vulnerabilities and agentic risk

Overview

This skill is coherent for agent-to-agent communication, but it broadly encourages external delegation and file transfer without clear per-use consent or data-minimization boundaries.

Review before installing. Use this only if you are comfortable with the agent contacting third-party agents and potentially sending prompts, metadata, and files. Require the agent to ask before each search, knock, message, file transfer, delegated task, detached host, policy change, or hook setup, and avoid sending secrets or regulated data unless you have explicitly approved the recipient and contents.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill promotes searching for other agents, delegating tasks, and sending files, but it does not prominently warn that doing so transmits user content to external agents and through network infrastructure such as relays. Even if native messages are end-to-end encrypted, disclosure still occurs to external recipients, and the privacy section explicitly notes that some bridge messages are readable by the relay, making omission of an upfront warning materially dangerous.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill encourages using Peerknock whenever the agent lacks knowledge, tools, data, or a second opinion, which is broad enough to trigger external delegation in many normal workflows. In a skill that can contact third-party agents and transfer files, vague invocation criteria increase the chance that sensitive user prompts, documents, or metadata are sent off-box without a deliberate trust check or user approval.

Content

No source excerpt is available for this finding.