The Agent Skills Directory

[SAFE]: No security threats or malicious patterns detected. The skill consists of Markdown-based guidelines and a project database for research purposes. It lacks any associated scripts or code that would allow for system modification or command execution.
[EXTERNAL_DOWNLOADS]: The skill references several external resources, including MIT's Density Atlas (densityatlas.org), UN-Habitat, and the Gehl Institute. While an automated scanner flagged one of these, analysis confirms they are legitimate academic and institutional domains used for research. These are static references and do not involve automated code downloads or remote execution.
[PROMPT_INJECTION]: The skill processes project-specific data provided by users or gathered from external research, representing a surface for indirect prompt injection. However, this risk is mitigated by the absence of any executable tools or scripts. The agent possesses no capabilities (such as file-system writes, network requests, or shell access) that could be leveraged maliciously through this ingestion surface.
Ingestion points: Precedent names and descriptions provided in user queries or retrieved by the agent.
Boundary markers: The skill enforces a structured five-dimension analysis framework for all outputs.
Capability inventory: No executable scripts, subprocess calls, or system tools are defined in the skill files.
Sanitization: While explicit input sanitization is not detailed, the lack of dangerous capabilities eliminates the possibility of exploitation via data ingestion.

precedent-study