Back to skill

Security audit

Gestalt

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed prompt-style collaboration skill with bounded memory guidance and no hidden execution, network access, credential handling, or destructive behavior.

Install this if you want the agent to answer more directly and, in coding-agent deployments, potentially keep repo-local notes for future sessions. Review the companion AGENTS.md/GEMINI.md behavior first if you do not want agents creating _memory, _tasks, or _comms files or relying on repository notes across sessions.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description says this skill changes conversational behavior for the session by applying Gestalt collaboration rules. The supplied code does not implement conversational mode changes, triggers, or session state. Instead, it is a local shell utility that reads AGENTS.md and writes GEMINI.md to mirror documentation content. That is a materially different primary purpose and involves undeclared file system read/write behavior unrelated to the described skill behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill is presented as a terse collaboration mode, but its instructions materially expand agent behavior into persistent repository memory, workflow management, and cross-session state handling. That mismatch can cause an agent to read, write, and rely on repo files beyond what a user would reasonably expect from a style-only mode, increasing the chance of unauthorized persistence, prompt-state contamination, or unintended disclosure through stored notes.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The repository-backed memory and inter-agent communication mechanisms are not necessary for delivering terse, substantive responses, so they introduce extra capability without clear justification. If followed, the agent may persist sensitive information, trust stale or malicious file content, or create covert communication/state channels that outlive the current interaction.

Content

No source excerpt is available for this finding.