100Critical
The skill enables arbitrary command execution via
Behavior Analysis
Description verified — behavior matches claim
Claims to do
Health: Audit the Six-Layer Stack: Prefix your first line with 🥷 inline, not as its own paragraph.
Actually does
This skill executes a shell script (`collect-data.sh`) to gather configuration data from the project's Claude Code setup, including `CLAUDE.md`, rules, skills, hooks, subagents, and verifiers. It performs a live check on configured MCP servers by calling one tool per server and accesses `settings.local.json`. Based on project complexity, it analyzes the data, potentially launching subagents (`inspector-context.md`, `inspector-control.md`) for deeper audits, and generates a structured health report with actionable fixes.
Skill Info
Nametw93/waza/health
Registrygithub
Versionb89a98a9e69e
PURLpkg:github/tw93/waza@b89a98a9e69e?skill=health
Stars4,497
Installs3,900
Source
SHA-25689903087db7a...
SHA-1e7f24290ff4a...
MD514cee951a12a...
TLSH95023b0b7d0a...
ssdeep96:UVM6yLVJw...
Size8,552 bytes
Install
Assessments (2)
AI Agent Traps ↗Arbitrary Command Execution via External Scriptcriticalcommand execution
The skill executes an external bash script (`collect-data.sh`) whose path can be influenced by environment variables or user-writable directories. A compromised script in these locations could lead to arbitrary command execution.
100% confidenceline 53
bash "$HEALTH_SCRIPT"; for candidate in ... "$HOME/.agents/skills/health/scripts/collect-data.sh"; do ... done
Sub-Agent Spawning Capabilityhighsub agent spawning
The skill is designed to launch two subagents in parallel for deep audits. This capability, if not properly constrained or if the subagent prompts are compromised, could lead to uncontrolled actions or resource abuse.
100% confidence
launch two subagents in parallel. ... Agent 1 (Context + Security): Read `agents/inspector-context.md`. ... Agent 2 (Control + Behavior): Read `agents/inspector-control.md`.
Suggestion of Executable Commands to Human Usermediumhuman social engineering
The skill's report includes an 'Action' field with copy-pasteable shell commands for remediation. If the skill's analysis logic is compromised, it could suggest malicious commands to the human user.
100% confidenceline 90
Action: <exact command or edit to fix>; Example: `git rm --cached .claude/settings.local.json && echo '.claude/settings.local.json' >> .gitignore`
Badge
Markdown
[](https://mondoo.com/ai-agent-security/skills/github/tw93/waza/health)HTML
<a href="https://mondoo.com/ai-agent-security/skills/github/tw93/waza/health"><img src="https://mondoo.com/ai-agent-security/api/badge/github/tw93/waza/health.svg" alt="Mondoo Skill Check" /></a>Image URL
https://mondoo.com/ai-agent-security/api/badge/github/tw93/waza/health.svgSecure your AI agents
Skills can read files, run commands, and access credentials. Mondoo helps organizations manage the security risks of AI agent skills across their entire fleet.
- Continuous scanning of skills across all registries
- Policy enforcement before skills reach your agents
- Integration with your existing security workflow