The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates for ingesting and processing vault data, which constitutes a surface for indirect prompt injection if the files contain malicious instructions.
Ingestion points: Vault markdown files are ingested via app.vault.read() in both SKILL.md (Step 2) and references/implementation-guide.md (ExportService, BackupService).
Boundary markers: Absent; the instructions do not implement or recommend delimiters to isolate untrusted file content during processing.
Capability inventory: The skill implements extensive file system operations including write, modify, create, and delete across SKILL.md and references/implementation-guide.md.
Sanitization: Absent; the provided logic processes vault content as raw text strings without validation or instruction-filtering mechanisms.
[DATA_EXFILTRATION]: The skill includes functionality to export aggregated vault data from the environment.
Evidence: The ExportService.downloadExport method in references/implementation-guide.md uses the URL.createObjectURL and a.click() pattern to trigger a browser-level download of vault data. This behavior is transparent and aligned with the skill's stated purpose of data export.
[SAFE]: The skill demonstrates best practices for Obsidian plugin development, such as utilizing the provided loadData and saveData methods for configuration management and leveraging metadataCache to avoid manual YAML parsing.

obsidian-data-handling