The Agent Skills Directory

[Prompt Injection] (SAFE): No patterns of instruction override, jailbreaking, or system prompt extraction were detected. The instructions are strictly focused on text transformation and slide formatting.
[Data Exposure & Exfiltration] (SAFE): No hardcoded secrets, sensitive file paths, or network communication protocols (like curl or fetch) are present.
[Obfuscation] (SAFE): The content is entirely human-readable with no Base64, zero-width characters, or hidden Unicode tags.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill contains a reference to npx skills add vercel-labs/agent-skills within a slide template. This is an example of generated text, not an instruction for the agent to execute code. The repository mentioned is also on the Trusted External Sources list.
[Indirect Prompt Injection] (LOW):
Ingestion points: The skill ingests user ideas and drafts via prompts like 'write content for...' or 'draft slides about...'.
Boundary markers: None present; the skill treats all input as content to be reformatted.
Capability inventory: No scripts or executable files are included in the skill; it has no capabilities to execute commands or write to the file system.
Sanitization: None; however, because the skill has no dangerous capabilities, the lack of sanitization does not pose a security risk.
[Dynamic Execution] (SAFE): No use of eval, exec, or runtime compilation. The skill only performs static text template substitution.

presentation-content