The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local script (look_tongji.py) located within the repository's directory structure to manage file imports. This is standard operational behavior for a tool-based skill.
[DATA_EXFILTRATION]: No network activity or unauthorized data transfer patterns were detected. The skill operates on local file paths provided during the workflow.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface as it processes external file formats (PDF, DOCX, PPTX).
Ingestion points: Files are provided via the --material flag in the CLI command described in SKILL.md.
Boundary markers: None explicitly defined in the instructions.
Capability inventory: File copying, Markdown conversion via MarkItDown, and wiki indexing.
Sanitization: None described within the skill instructions. While this represents a theoretical attack surface, it is consistent with the primary purpose of the skill and does not indicate malicious intent.

add