The Agent Skills Directory

[SAFE]: The skill is a specialized text editor that uses internal markdown documents for its logic. Analysis of the instructions and reference files reveals no malicious patterns, external dependencies, or attempts to bypass security controls. It does not perform network operations or access sensitive system files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided text, which is an untrusted data source. While this represents a potential injection surface, the risk is assessed as safe because the skill is not equipped with any tools that would allow an injected instruction to perform harmful actions, such as exfiltrating data or modifying the environment. The multi-stage editing process specifically targets and removes meta-commentary and redundant rhetorical structures, which provides a layer of semantic filtering. 1. Ingestion points: User-provided text for polishing. 2. Boundary markers: The skill uses distinct output headers like '=== 修改后文案 ===' to separate results. 3. Capability inventory: None detected; no file-write, network, or shell execution tools are used. 4. Sanitization: The '去 AI 感' (Remove AI markers) stage explicitly removes meta-instructions and filler text.

text-polisher