The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local scripts, specifically node scripts/new-episode.mjs <slug> and bash scripts/check-episode-contract.sh, where the <slug> argument is derived directly from user input (e.g., when the user requests to "新建一期"). This creates a potential command injection surface if these scripts do not implement strict input validation.
[INDIRECT_PROMPT_INJECTION]: The skill processes external content which can be manipulated to include hidden instructions.
Ingestion points: The agent reads from 01-口播稿.md (video scripts) and 07-剪辑/subtitles.srt (subtitles) as described in the workflow instructions.
Boundary markers: There are no mentioned boundary markers or explicit instructions to treat these file contents as inert data, making the agent susceptible to instructions embedded within the content.
Capability inventory: The skill has permissions to write to project state files (e.g., 09-导演/director-state.md) and execute shell commands via Node.js and Bash.
Sanitization: There is no evidence of sanitization or filtering of the text content before it is processed by the agent.
[DYNAMIC_EXECUTION]: The skill relies on the execution of external .mjs and .sh scripts at runtime to manage project lifecycle and configuration validation. Since the contents of these scripts (e.g., new-episode.mjs and check-episode-contract.sh) are not provided in the skill package, their interaction with the host system is opaque.

laohan-daoyan