disler-observability-hooks

The skill's hooks execute bash commands that incorporate user-provided content without proper sanitization. Multiple hook scripts use subprocess.run() and similar methods to execute shell commands where user input could be interpolated. The most concerning example is in the TTS functionality where user messages are passed directly to shell commands. Additionally, the pre_tool_use.py script processes bash commands from user input and could allow injection of additional shell metacharacters or command chaining.

The skill's hook scripts contain multiple instances of bash commands that use force flags to bypass safety checks and confirmations. Specifically, the pre_tool_use.py script checks for dangerous rm commands but allows them in certain directories, and multiple scripts execute system commands without user confirmation. The notification.py script executes TTS commands with subprocess calls that could potentially be exploited.

The skill's hook scripts access environment variables (ELEVENLABS_API_KEY, OPENAI_API_KEY, ENGINEER_NAME) and execute system-level commands through subprocess calls. The notification.py script determines which TTS system to use based on available API keys and executes external scripts. The pre_compact.py script loads environment variables using dotenv and can backup files to system directories.

The skill includes multiple hooks that call 'send_event.py' with various event types and data summarization flags. This script is referenced extensively but not provided in the skill content, making it impossible to verify where this data is being sent. The --source-app parameter suggests data is being transmitted to an external observability system.

The skill contains multiple conditional branches that depend on environment variables (ELEVENLABS_API_KEY, OPENAI_API_KEY, ENGINEER_NAME) and system state (file existence, random number generation) that users would not anticipate. The TTS selection logic chooses different audio providers based on API keys, and notification behavior varies based on engineer name presence and random chance. These conditionals are not documented in the skill's behavioral guarantees.

The skill declares empty scope and dependencies but contains 11 substantial companion scripts (notification.py, permission_request.py, post_tool_use.py, etc.) totaling thousands of lines of code. These scripts perform significant functionality including TTS audio generation, file system operations, transcript backups, and external API calls that are not declared in the skill metadata.

The skill executes Python scripts using 'uv run' commands that contain dependency declarations in script headers (/// script blocks). These dependencies include python-dotenv and other packages that are not listed in the skill's declared dependencies field, which is empty. This creates implicit package installations that bypass dependency review.