The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run to interact with the Thinkbox Deadline deadlinecommand CLI. The path to the executable can be optionally provided as a tool argument (deadline_command), which allows the execution of arbitrary binaries on the host system. This pattern is present across multiple scripts including cancel_render_job.py, get_render_job_status.py, list_render_jobs.py, render_farm_status.py, and submit_render_job.py.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the active Blender scene which is then used in downstream operations.
Ingestion points: Scene file paths, job names, and render settings are retrieved via the bpy module in submit_render_job.py, validate_scene_for_farm.py, and write_render_job.py.
Boundary markers: None identified in the prompt templates or data processing logic to separate scene data from instructions.
Capability inventory: The skill has the capability to execute shell commands via subprocess.run, write files to disk via open().write(), and perform network operations via urllib.request.
Sanitization: The skill uses list-based arguments for subprocess calls and json.dumps() for API payloads, which provides some protection against basic injection, but does not prevent logic-based attacks via malicious scene metadata.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to Flamenco Manager instances to manage render jobs. The flamenco_server_url is configurable via tool arguments or environment variables (FLAMENCO_SERVER_URL), directing the agent to communicate with external endpoints. This network activity is used in cancel_render_job.py, get_render_job_status.py, list_render_jobs.py, render_farm_status.py, and submit_render_job.py.

blender-render-farm