The Agent Skills Directory

[Indirect Prompt Injection] (SAFE): Evaluated the skill for potential exploitation via untrusted data ingestion.
Ingestion points: The 'Create your Own Theme' section in SKILL.md allows the agent to process user-provided descriptions to generate new themes.
Boundary markers: Absent. The instructions do not explicitly tell the agent to ignore instructions embedded within the user's theme descriptions.
Capability inventory: The skill is limited to reading from the local themes/ directory and applying fonts and colors to artifacts. It lacks capabilities for command execution, network access, or sensitive file manipulation.
Sanitization: No sanitization logic is described for user inputs.
Conclusion: While an ingestion surface exists, the lack of powerful capabilities makes this surface low-risk.
[Data Exposure & Exfiltration] (SAFE): No sensitive file paths (e.g., SSH keys, credentials) or network exfiltration patterns were detected. The skill only accesses a specific local themes/ directory and a theme-showcase.pdf file.
[Remote Code Execution] (SAFE): No patterns for downloading and executing remote scripts or packages were found.

theme-factory