100Critical
This skill generates and executes arbitrary bash scripts
Behavior Analysis
Description verified — behavior matches claim
Claims to do
Task Orchestrator Hooks Builder Skill: You are a hook automation specialist helping users create hooks that integrate Task Orchestrator's workflow cascade event system with their workflow. This skill works with any MCP client that supports hooks (Claude Code, Claude Desktop, Cursor, Windsurf, etc.).
Actually does
This skill uses `Read`, `Write`, and `Bash` tools to generate, configure, and manage shell scripts (`.sh` files) that act as 'hooks' for the Task Orchestrator system. These hooks parse JSON input (from stdin) using `jq`, interact with a local SQLite database (`tasks.db`), execute `git` commands, and can potentially make `curl` requests to external URLs, all within the `$CLAUDE_PROJECT_DIR`.
Skill Info
NameAngriff36/Capsule/task-orchestrator-hooks-builder
Registrygithub
Versiona7b7968a168f
PURLpkg:github/Angriff36/Capsule@a7b7968a168f#task-orchestrator-hooks-builder
Source
SHA-256cf10c4d11017...
SHA-18d11add4cefe...
MD5d613785ffc25...
TLSHaf621a26f955...
ssdeep192:uq+QMHEl...
Size14,877 bytes
Install
This skill has been flagged as potentially malicious. Review the findings below before installing.
Assessments (2)
AI Agent Traps ↗Shell command execution function detectedcriticalcommand execution
Shell command execution function detected
100% confidenceline 47
system (
Arbitrary Command Execution via Bash Scriptscriticalcommand execution
The skill's primary function is to generate and execute arbitrary bash scripts based on user input. This capability allows for direct command execution on the host system, posing a significant risk if malicious script content is introduced.
100% confidenceline 5
allowed-tools: Bash, 'Perform the action # ... your automation logic here ...', 'Test execution/quality gates: ./gradlew test'
Persistent Code Execution via Hook Configurationhighpersistence
The skill creates and modifies executable bash scripts in `.claude/hooks/` and updates `.claude/settings.local.json` to configure these hooks. This allows for persistent execution of attacker-controlled code whenever the defined trigger conditions are met.
100% confidenceline 100
'Configuring settings - Add hook to .claude/settings.local.json', 'Create Hook File: Write hook script to .claude/hooks/[descriptive-name].sh', 'Make script executable'
Data Exfiltration and External Communicationhighdata exfiltration
The skill explicitly supports 'External API Integration' using `curl` within generated bash scripts. This enables scripts to make arbitrary network requests, potentially exfiltrating sensitive data from the agent's environment or the user's system to external servers.
100% confidenceline 370
'External API Integration: curl -X POST https://api.example.com/notify -H "Content-Type: application/json" -d "{\"task_id\": \"$TASK_ID\", \"status\": \"$STATUS\"}"'Resource Abuse Potential via Script Executionmediumresource abuse
The ability to generate and execute arbitrary bash scripts introduces the risk of resource abuse, such as consuming excessive CPU, memory, or disk space through maliciously crafted scripts (e.g., infinite loops, fork bombs).
90% confidenceline 80
'Perform the action # ... your automation logic here ...', 'Keep hooks fast - long-running hooks slow Claude's workflow'
RAG Poisoning via Generated Documentationlowrag poisoning
The skill instructs the agent to generate documentation (`README.md`) for the created hooks. If this documentation is later indexed by the agent's RAG system, it could be used to inject misleading or harmful information into the agent's knowledge base.
70% confidenceline 200
'Add documentation to .claude/hooks/README.md'
Badge
Markdown
[](https://mondoo.com/ai-agent-security/skills/github/Angriff36/Capsule/task-orchestrator-hooks-builder)HTML
<a href="https://mondoo.com/ai-agent-security/skills/github/Angriff36/Capsule/task-orchestrator-hooks-builder"><img src="https://mondoo.com/ai-agent-security/api/badge/github/Angriff36/Capsule/task-orchestrator-hooks-builder.svg" alt="Mondoo Skill Check" /></a>Image URL
https://mondoo.com/ai-agent-security/api/badge/github/Angriff36/Capsule/task-orchestrator-hooks-builder.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