Back to skill

Security audit

create-team

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed template for building Claude Code agent-team packages, with no evidence of hidden execution, data theft, or destructive behavior.

Install or run generated teams only when you are comfortable with Claude Code spawning multiple teammates and with any generated agent definitions or hooks being added to .claude locations. Prefer project-scoped .claude/agents setup unless you intentionally want reusable user-wide agents, and review generated tools allowlists and hook commands before enabling them.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · SKILL.md (reported line 85)May include surrounding context.

md
Keep them in sync — drift between `agents/` and `references/` is the #1 failure mode.

### 4. Write SKILL.md as a lead playbook

Required sections (skeleton in [references/templates.md](references/templates.md)):

Agent Config Directory Access

High
Category
Agent Snooping
Confidence
90% confidence
Finding

Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.

Content

Scanner excerpt · references/templates.md (reported line 42)May include surrounding context.

md
- Verify Claude Code version: `claude --version` ≥ 2.1.32
- Verify env: `echo $CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` returns `1`
- Choose `teammateMode` in `~/.claude/settings.json` (`in-process` or `tmux`)

## Install Agents (one-time per project or user)

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 9)May include surrounding context.

md
language: en
---

# Create A Runnable Agent Team Skill

## When To Use

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 · references/templates.md (reported line 50)May include surrounding context.

md
\```bash
# project scope
mkdir -p .claude/agents
cp {SKILL_ROOT}/agents/*.md .claude/agents/

# OR user scope (reusable across projects)

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

The frontmatter sets language: en, which is a natural-language locale restriction. The file does not offer user opt-in, alternatives, or a documented region-specific justification, so this appears to force a specific language policy unnecessarily.

Content

No source excerpt is available for this finding.