The Agent Skills Directory

[PROMPT_INJECTION]: The documentation describes features that enable the ingestion of untrusted data (via tp.system.prompt and tp.obsidian.request) and its interpolation into templates, creating a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the context through tp.system.prompt(), tp.system.suggester(), and network responses from tp.obsidian.request() as documented in SKILL.md.
Boundary markers: The documentation does not specify the use of delimiters or specific instructions to the agent to ignore embedded commands within processed data.
Capability inventory: The skill documents powerful capabilities including system command execution (tp.system.executeCommandWithOutput()), file creation (tp.file.create_new()), and HTTP networking (tp.obsidian.request()) in SKILL.md.
Sanitization: The documentation provides a dedicated 'Security Warning' section advising users to review code, trust only known templates, and notes that system commands are restricted in modern sandbox environments.
[COMMAND_EXECUTION]: The skill provides instructions for using tp.system.executeCommandWithOutput() to run arbitrary shell commands. Although it contains warnings about plugin-level restrictions and sandbox limitations, this functionality allows an agent to interact with the host system's command line.
[DATA_EXFILTRATION]: Documentation outlines the use of tp.obsidian.request() and tp.obsidian.requestUrl() for network operations. While these are legitimate plugin features, they could be misused to transmit vault data or local file contents to external servers.
[NO_CODE]: The skill consists exclusively of documentation and markdown instructions; it does not contain or execute any standalone scripts or binary files.

obsidian-plugin-templater