The Agent Skills Directory

[SAFE]: The skill demonstrates legitimate research functionality with no evidence of malicious intent, prompt injection, or data exfiltration. All external tools are used through a controlled gateway with read-only guidelines.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, which presents a surface for indirect prompt injection.
Ingestion points: Data is ingested via exa_search, tavily_search, and wechat_mp_v2_account_articles (SKILL.md).
Boundary markers: The instructions explicitly direct the agent to "Separate factual observations from analysis and recommendations" and "cite sources" to maintain data integrity.
Capability inventory: The skill is restricted to "read-only research" and synthesizing findings; it lacks capabilities for persistent storage, file system modification, or arbitrary network access.
Sanitization: There are no explicit sanitization or filtering protocols mentioned for the content retrieved from external sources.

newsletter-research