The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified. The skill consists entirely of instructional markdown and reference materials designed to guide an AI agent in writing balanced political news.
[PROMPT_INJECTION]: The skill does not contain any instructions aimed at bypassing safety filters or overriding system prompts. It uses instructional language to enforce journalistic standards rather than to subvert agent behavior.
[DATA_EXFILTRATION]: There are no network operations, hardcoded credentials, or attempts to access sensitive system files. All referenced URLs are to legitimate government, educational, or news organization domains (e.g., .gov.tw, .edu).
[REMOTE_CODE_EXECUTION]: The skill does not define any external dependencies, package installations, or remote script downloads. It operates purely as a set of logical instructions for text processing.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process untrusted user material (transcripts, leaked documents), it specifically mandates a 'material audit' and 'fact-check' process via its parent skill (med-news-reporter) and includes a 'Frame Neutrality Audit' to mitigate risks associated with biased or malicious input data.

med-political