The Agent Skills Directory

[SAFE]: The skill operates entirely within the local file system, specifically restricted to the workspace/ directory. It does not perform any network operations, credential harvesting, or access to sensitive system configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as video titles and script files. The potential risk of instructions being embedded in these files is mitigated by the skill's hard rules, which include disable-model-invocation: true and a requirement for explicit user invocation via the /ship-episode command. This ensures the agent cannot be autonomously triggered by malicious content.
[COMMAND_EXECUTION]: The skill utilizes local scripts (helpers/loudness.py and hooks/validate-export-name.js) for validation. These tools are invoked via the authorized Bash environment to perform specific project tasks and do not involve dynamic code generation or remote content execution.

publish-operator