The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were identified. The skill defines administrative tasks for research projects and uses standard agent patterns for state management.- [PROMPT_INJECTION]: The skill features an indirect prompt injection surface as it reads project context from .memory/activeContext.md. 1. Ingestion points: projects/{slug}/.memory/activeContext.md. 2. Boundary markers: Absent. 3. Capability inventory: create_project, switch_project, update_project_settings. 4. Sanitization: Absent. This surface is considered safe given it is restricted to the skill's operational directory and is inherent to its primary purpose.- [DATA_EXFILTRATION]: The skill interacts with local files in the projects/ directory to persist progress. No access to sensitive system paths or unauthorized network operations were detected.

project-management