The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection attempts, safety filter bypasses, or instructions to ignore previous rules were detected. The content is purely instructional for plugin development.
[DATA_EXFILTRATION]: No hardcoded credentials, sensitive file access outside the application context, or suspicious network operations to non-whitelisted domains were found. The code snippets use standard Obsidian API calls for vault operations.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known development tools and libraries (esbuild, obsidian, typescript) which are appropriate for the stated purpose of plugin development.
[COMMAND_EXECUTION]: While the guide mentions build commands (npm run dev/build), these are standard development workflows and do not involve arbitrary or malicious command injection.
[OBFUSCATION]: No obfuscated code, Base64 strings, zero-width characters, or hidden payloads were identified.
[INDIRECT_PROMPT_INJECTION]: The skill serves as a template for generating code. While code generation tasks inherently carry a surface for indirect prompt injection if the agent processes untrusted user input, this specific skill provides static, safe boilerplate templates.
Ingestion points: User-provided plugin names or logic descriptions in SKILL.md template sections.
Boundary markers: Not explicitly defined in templates.
Capability inventory: File reading/writing within Obsidian vault via this.app.vault.
Sanitization: Standard template-based approach; the agent should ensure user input does not inject malicious logic into the final plugin code.

obsidian-plugin-dev