Back to skill

Security audit

internal-comms

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent internal communications helper, but it tells the agent to broadly search company Slack, email, documents, and calendars without clear privacy limits.

Install only if you are comfortable with the agent using connected company systems for internal communications. Before use, explicitly limit sources to approved channels, documents, and date ranges; avoid private inboxes, restricted channels, HR, legal, security, finance, personnel, customer, or deal-sensitive material unless specifically authorized; and review outputs before sharing broadly.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill description says it should be used whenever asked to write some sort of internal communications, which is a broad activation scope. Overly broad routing can cause the agent to invoke this skill for loosely related requests, increasing the chance of inappropriate context loading or unintended instruction influence from auxiliary files.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The keyword list includes vague terms such as 'updates' and 'weekly update' that are common across many unrelated tasks. Generic trigger terms can cause accidental invocation, making the skill apply internal communication guidance in contexts where it does not belong and expanding exposure to any risky instructions in referenced resources.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly directs the agent to collect information from Slack, Google Drive, Email, and Calendar, including signals like channel size, reactions, views, and response volume, but provides no constraints on authorization, minimum necessary access, or handling of sensitive/internal data. In an internal communications context, this can cause over-collection of confidential business, personnel, or customer information and inclusion of data from sources the user did not intend to authorize for this task.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly encourages broad access to internal Slack, email, calendar, documents, and external press sources to assemble a company-wide newsletter, but it provides no guidance on filtering for confidentiality, minimizing data collection, or obtaining user confirmation before using sensitive sources. In a large company context, this creates a real risk of exposing private communications, confidential documents, executive-only announcements, or meeting materials that were not intended for broad redistribution.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The instructions expand the skill from formatting internal communications into proactively collecting and synthesizing information from broad company systems. That scope expansion can cause unnecessary access to sensitive internal discussions and creates a data-collection behavior not clearly justified by the stated skill purpose.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill directs the agent to survey Slack, email, and documents across the company to identify common confusion topics, which is broader than necessary for a communications-writing helper. This increases the chance of over-collection, exposure of confidential information, and use of employee communications outside their original context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instructions encourage broad review of Slack, email, and documents without any privacy, confidentiality, or data-minimization warnings. In a company environment, those systems often contain HR, legal, security, or personal data, so unrestricted review can expose sensitive information during summarization or retrieval.

Content

No source excerpt is available for this finding.