100Critical
The skill allows arbitrary command execution via `
Behavior Analysis
Description verified — behavior matches claim
Claims to do
Context Search (QMD): Search past Claude Code sessions and knowledge bases using QMD.
Actually does
The skill executes a Python script (`search-tool.py`) via `uv run` in a Bash shell. This script reads `search-config.yaml` from the current working directory to search configured collections using an underlying `qmd` tool. It outputs formatted search results including file paths and line numbers.
Skill Info
Namevranac/claude-context-search-qmd/context-search
Registrygithub
Versiona2aceafd1859
PURLpkg:github/vranac/claude-context-search-qmd@a2aceafd1859?skill=context-search
Source
SHA-256fcab1ba2d9e2...
SHA-1dd2410390142...
MD5986f22bcba22...
TLSH175154577104...
Size2,475 bytes
Install
This skill has been flagged as potentially malicious. Review the findings below before installing.
Assessments (2)
AI Agent Traps ↗Unrestricted `uv run` command executioncriticalcommand execution
The `allowed-tools` configuration permits `uv run` with arbitrary arguments (`uv run:*`). This allows an attacker to execute any command or script via `uv run`, potentially leading to full system compromise.
95% confidenceline 3
allowed-tools: "Bash(uv run:*)"
Potential command injection in search querymediumcommand execution
The 'your search query' argument passed to `search-tool.py` might be vulnerable to command injection if the script uses it unsanitized in a shell command or external process call.
70% confidenceline 19
uv run ${CLAUDE_PLUGIN_ROOT}/skills/context-search/scripts/search-tool.py "your search query"Malicious configuration injection via `--cwd`highrag poisoning
The `search-tool.py` script reads `search-config.yaml` from the current working directory, which can be controlled by the `--cwd` option. An attacker could inject a malicious configuration file to manipulate search results, exfiltrate data, or trigger unintended actions.
90% confidenceline 19
uv run ... --cwd /path The tool reads `search-config.yaml` from the working directory
Badge
Markdown
[](https://mondoo.com/ai-agent-security/skills/github/vranac/claude-context-search-qmd/context-search)HTML
<a href="https://mondoo.com/ai-agent-security/skills/github/vranac/claude-context-search-qmd/context-search"><img src="https://mondoo.com/ai-agent-security/api/badge/github/vranac/claude-context-search-qmd/context-search.svg" alt="Mondoo Skill Check" /></a>Image URL
https://mondoo.com/ai-agent-security/api/badge/github/vranac/claude-context-search-qmd/context-search.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