The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script for data validation.
Evidence: In SKILL.md, the agent is told to run python3 scripts/check-publish-format.py <发布资料文件> after saving the generated file.
Context: This is used to ensure the generated publishing materials adhere to the specific formatting rules of the target platform (e.g., Fanqie).
[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated content from various project files, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads data from files such as 设定/题材定位.md, 大纲/大纲.md, 设定/角色/, 设定.md, and 小节大纲.md.
Boundary markers: Absent. There are no explicit instructions or delimiters used to separate the untrusted project data from the agent's instructions.
Capability inventory: The agent can write files to the local system (发布资料-<平台>.md) and execute the scripts/check-publish-format.py script.
Sanitization: The scripts/check-publish-format.py script provides a layer of sanitization by validating the format and content of the output file against platform references.
[DYNAMIC_EXECUTION]: The skill utilizes a local script for runtime processing of data.
Evidence: The skill workflow involves executing scripts/check-publish-format.py at runtime to verify the generated publishing document.
Context: This is a low-risk implementation as the script is distributed with the skill and performs read-only validation logic.

story-publish