The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the user's local vault to perform the audit, creating a risk if malicious instructions are embedded in the notes.
Ingestion points: Vault notes and the AGENTS.md configuration file are read and processed as described in SKILL.md (Step 1 and 3).
Boundary markers: The instructions do not establish clear delimiters or protective prompts to distinguish between vault data and operational instructions during the audit process.
Capability inventory: The skill includes file system modification capabilities ('safe-write' rules) referenced in SKILL.md (Step 5) to apply fixes after user approval.
Sanitization: No sanitization or filtering of the vault's markdown content is mentioned prior to the analysis and recommendation phase.
[COMMAND_EXECUTION]: The skill utilizes a local command-line interface to interact with the vault.
Evidence: SKILL.md (Step 2) instructs the agent to execute obsidian help <command> to verify CLI syntax and inspect the current vault state.

vault-audit