obsidian-vault
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use shell utilities like
findandgrepto perform file searches and content indexing within the path/mnt/d/Obsidian Vault/AI Research/. These are standard tools used for local file management. - [PROMPT_INJECTION]: As the skill is designed to read and process content from user-generated markdown files, it possesses an attack surface for indirect prompt injection. Malicious instructions placed inside a vault note could potentially be executed by the agent when it searches or reads the note content.
- Ingestion points: Markdown (.md) files located in the hardcoded vault directory.
- Boundary markers: Not implemented; the agent reads raw file content via grep.
- Capability inventory: Subprocess execution (find, grep), file system navigation, and file creation/modification.
- Sanitization: No explicit sanitization or escaping of note content is defined before processing.
Audit Metadata