The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill features a tool named generate_blender_script that accepts a user-provided ollama_url to fetch and execute dynamically generated Python scripts within the Blender process.
[REMOTE_CODE_EXECUTION]: The manage_blender_addons tool allows the installation of Blender addons, which are Python-based extensions that execute with the same privileges as the Blender application.
[EXTERNAL_DOWNLOADS]: Through the download_and_import tool, the skill can fetch files from arbitrary URLs. This presents a risk of ingesting malicious assets that may trigger auto-run scripts or exploit vulnerabilities in file parsers.
[DATA_EXFILTRATION]: Tools such as blender_export and manage_object_repo enable the skill to write files to arbitrary local paths, which can be leveraged to move sensitive data to publicly accessible directories or networked drives.
[COMMAND_EXECUTION]: The skill's primary function includes the generation and execution of Python scripts to automate 3D modeling tasks, effectively providing a scriptable interface to the underlying operating system via Blender's API.
[PROMPT_INJECTION]: The skill exposes a significant surface for indirect prompt injection through tools that process natural language for autonomous workflows.
Ingestion points: The workflow_prompt parameter in agentic_blender_workflow and the prompt parameter in intelligent_3d_processing ingest untrusted text.
Boundary markers: No delimiters or safety instructions are defined to separate user input from system instructions.
Capability inventory: The agent possesses powerful capabilities including file writing (blender_export), script generation (generate_blender_script), and system configuration modification (config_set).
Sanitization: There is no evidence of sanitization or validation of the input prompts before they are used to drive multi-step workflows.

blender-expert