The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/tzfilm_douyin_adapter.py uses runpy.run_path to execute a Python script located at a hardcoded path in the user's home directory (~/Documents/laohanAI视频创作/scripts/tzfilm_douyin_adapter.py). This allows for the execution of arbitrary code that is not contained within the skill itself, creating a dependency on external, unmanaged code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to collect and aggregate user comments from Douyin (TikTok) and provide them to an upstream 'retro' engine (cheat-retro) for analysis.
Ingestion points: Raw comment text is written to 14-评论/comments.jsonl from external sources (Douyin session or manual paste).
Boundary markers: The instructions do not specify the use of XML tags, delimiters, or sanitization techniques to prevent malicious instructions within comments from influencing the agent during the 'retro' phase.
Capability inventory: The skill has the ability to write to the file system and execute shell commands (python3, node).
Sanitization: There is no mention of escaping or filtering content beyond deduplication and empty text removal.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to execute multiple shell commands, including node scripts/register-cheat-publish-evidence.mjs and python3 scripts/tzfilm_douyin_adapter.py with various arguments. While these appear intended for the workflow, they represent a significant capability surface area.

laohan-yunying