The Agent Skills Directory

[SAFE]: The skill instructions define a benign workflow for product strategy and research. No patterns associated with privilege escalation, persistence, or credential harvesting were found.- [DATA_EXFILTRATION]: The skill guides the agent to use web search and fetch tools during the research phase. This network access is directed at public information gathering, such as market data and competitor positioning, and does not involve exfiltration of private user data or sensitive system files.- [PROMPT_INJECTION]: The skill contains internal logic to enforce a specific workflow order, such as requiring research to be completed before a Product Requirements Document is generated. These are functional instructions for the agent and do not contain patterns intended to bypass AI safety guidelines or extract system prompts.

idea-os