The Agent Skills Directory

[DATA_EXPOSURE_&_EXFILTRATION]: The skill implements a persistent state system using markdown files in the local workspace. It creates and updates files within a structured directory (e.g., 0-原石/0.5-体系学习/) to store learning maps, knowledge gaps, and lesson history. The skill explicitly limits its scope to user-authorized files and directories, stating it must not silently scan the workspace.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted user data, such as articles and notes, to identify knowledge gaps. This input surface could theoretically be used to embed malicious instructions. The skill mitigates this risk by maintaining a strict separation between user-provided 'corpus' and its own internal state tracking and interpretation logic.
[COMMAND_EXECUTION]: While the skill manages a complex workflow, it does not utilize shell commands, subprocesses, or dynamic code execution. All logic is contained within the prompt instructions and structured file operations.

cuihuo-learning