The Agent Skills Directory

[SAFE]: No malicious behavior detected. The skill focuses on workflow management, state authority validation, and consistency checking within a business discovery context.
[PROMPT_INJECTION]: The skill includes strong directives (e.g., "MUST verify", "Immediately yield control") to prioritize the exploration-cycle state over freeform conversation. These instructions are operational and do not target safety filter bypasses or behavior override. The agent also processes external phase artifacts, which is an indirect injection surface.
Ingestion points: exploration/exploration-dashboard.md and artifacts like discovery plans, visual layouts, and prototype READMEs mentioned in SKILL.md.
Boundary markers: Absent; the agent is instructed to read content directly.
Capability inventory: File reading via the Read tool and invocation of the exploration-workflow skill.
Sanitization: None; the agent is instead tasked with validating consistency and scanning for unresolved placeholders manually.
[DATA_EXFILTRATION]: The skill reads from a local markdown file and a SQLite database to maintain state. There are no network operations, hardcoded credentials, or patterns suggesting data exfiltration.

using-exploration-cycle