The Agent Skills Directory

[SAFE]: The skill functions as a professional content review入口, managing historical records and generating testable content hypotheses without performing unauthorized network operations or privilege escalation.
[DATA_EXPOSURE]: The skill is designed to maintain a local record library (./eva-review/). It explicitly instructs the agent to avoid copying private information from comments or screenshots into shared cards, ensuring that only the minimum necessary metadata is stored for analysis.
[DYNAMIC_EXECUTION]: While the skill dynamically generates file paths based on platform and account names (e.g., ./eva-review/accounts/{platform}__{account}/), it incorporates a specific security protocol requiring the purification of path characters to prevent directory traversal attacks.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external content (Excel, CSV, screenshots, and comments). It mitigates this attack surface by referencing external safety boundaries and using structured markdown templates to contain and verify data before it influences subsequent agent actions.

eva-review