The Agent Skills Directory

[SAFE]: The skill implements a secure workflow for text generation. It guides the agent to gather topic details from the user and conduct web searches for competitive research. All identified operations are aligned with the stated purpose of writing tweets. There are no attempts to access sensitive system information, bypass agent constraints, or perform unauthorized network activity.\n- [INDIRECT_PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection as it processes external web search data (SKILL.md, Workflow Step 1) to influence the drafting of tweets. However, the capability tier is restricted to text drafting for human review, and the skill includes quality bars to ensure that metrics and results are sourced correctly from the user or cited research, mitigating the impact of potentially poisoned external content.\n- [DATA_EXFILTRATION]: The skill's data handling is limited to user-provided content and public web research. It does not access restricted file paths, credentials, or environment variables, and it contains no hardcoded secrets or unauthorized data transmission patterns.

tweet-writer