The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (look_tongji.py) via the shell to build and serve a static site. The script is accessed using a relative path (../../scripts/) within the repository, which is standard for local tool integration.
[PROMPT_INJECTION]: The skill identifies a data ingestion surface where it processes workspace data (lecture sessions, notes, and transcripts) to generate the wiki content. While this creates a potential surface for indirect prompt injection if the input data is untrusted, the skill's primary function is local serving and site generation, posing no immediate threat to the agent's core instructions in this context.
[EXTERNAL_DOWNLOADS]: The skill references a vendored frontend ('llm-wiki'). No remote downloads or external fetches were detected in the skill's workflow; the frontend is treated as a local asset.

wiki