The Agent Skills Directory

[DATA_EXFILTRATION]: The skill documentation includes tools for accessing sensitive user data, specifically the system clipboard (tp.system.clipboard()) and full file content (tp.file.content). It also details the tp.web.request() function for making outbound HTTP requests. The combination of these capabilities creates a potential data exfiltration vector where local information could be sent to an external server.
[DYNAMIC_EXECUTION]: The skill facilitates the creation of templates containing executable JavaScript blocks (<%* ... %>) and provides access to local user-defined scripts (tp.user.myFunction), which allows for dynamic code execution within the Obsidian application environment.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by enabling the ingestion of untrusted data from the filesystem or clipboard.
Ingestion points: SKILL.md (documentation for tp.file.content and tp.system.clipboard()).
Boundary markers: Absent; the documentation does not suggest using delimiters or specific instructions to sanitize ingested content.
Capability inventory: File system modification (tp.file.create_new, tp.file.rename, tp.app.vault.append) and network operations (tp.web.request).
Sanitization: Absent; no sanitization or validation methods are proposed for processing external data.

templater