ai-memory-obsidian
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to have the agent 'Check Memory First' by reading from external Obsidian vault notes and local memory indexes. This creates a surface for indirect prompt injection where instructions stored in the vault could influence agent behavior.
- Ingestion Points: Obsidian vault Markdown files and the ai-memory SQLite database index.
- Boundary Markers: No specific delimiters or markers are defined to isolate retrieved memory context from system instructions.
- Capability Inventory: The skill facilitates reading and writing local files and executing memory-related CLI tools.
- Sanitization: The skill mandates redacting sensitive data before writing to the vault, though it does not specify sanitization for data retrieved from the vault.
- [EXTERNAL_DOWNLOADS]: The skill documentation includes commands for installing the skill and its associated integrations from public registries and repositories.
- Fetches skill components from the author's GitHub repository using npx.
- Recommends the @bitbonsai/mcpvault package from the NPM registry for vault integration.
- References the ai-memory daemon repository on GitHub for optional local indexing.
- [COMMAND_EXECUTION]: The skill instructions guide the agent in using local command-line tools to interact with the memory services.
- Details the use of the official Obsidian CLI for searching and opening notes.
- Describes executing the ai-memory bridge command to facilitate communication between the agent and the local memory daemon.
Audit Metadata