The Agent Skills Directory

[SAFE]: No malicious patterns such as prompt injection, obfuscation, or credential harvesting were detected. The skill's logic is transparent and focuses on content analysis and strategy.
[COMMAND_EXECUTION]: The skill is designed to automatically write full analysis reports to the user's local directory at 01-内容生产/待深化的选题/. This is a core, documented feature of the skill intended for document management and persistent storage of AI-generated insights.
[PROMPT_INJECTION]: The skill ingests untrusted user data (topic ideas) for processing. While it does not employ explicit boundary markers or delimiters for this data, the risk of indirect prompt injection is mitigated by the skill's multi-stage analytical process, which transforms the input through specific diagnostic frameworks rather than executing it as commands.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads from a related local skill file (li-writer/SKILL.md) to align with the user's writing style. This cross-skill interaction is used for configuration and personalization and does not involve exfiltrating data to external domains.

li-topic