The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes note content from an Obsidian vault, which serves as untrusted data. If a note contains malicious instructions, the agent might follow them during processing. The skill lacks explicit boundary markers or instructions for the agent to ignore instructions embedded in the note data.
Ingestion points: Vault data is retrieved via obsidian base:query in SKILL.md.
Boundary markers: The instructions do not define delimiters or specific 'ignore' prompts for the ingested note content.
Capability inventory: The agent has capabilities to read, search (Grep, Glob), and modify (obsidian append) files within the vault.
Sanitization: While the skill correctly advises using shell variables to prevent command injection, it does not provide methods for sanitizing the natural language content of the notes against prompt injection.

bases