mempalace
Pass
Audited by Gen Agent Trust Hub on Apr 20, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill implements a Retrieval-Augmented Generation (RAG) architecture that indexes untrusted external data, creating a surface for indirect prompt injection. * Ingestion points: Local conversation exports (JSON, JSONL, TXT) and project files indexed via the 'mempalace mine' command. * Boundary markers: Documentation does not define specific delimiters or instructions to ignore embedded commands in retrieved 'drawers'. * Capability inventory: Includes 19 MCP tools for searching, reading, and modifying the palace memory and knowledge graph. * Sanitization: No mentioned validation or sanitization of retrieved verbatim text before it is presented to the agent.
- [COMMAND_EXECUTION]: The skill requires the installation of a CLI package and execution of shell commands ('mempalace mine', 'mempalace watch') for operations. It also suggests the use of a pre-compact shell hook script within the agent's environment.
- [EXTERNAL_DOWNLOADS]: The system fetches approximately 500 MB of embedding model data from established public repositories upon initial setup to support local search functionality.
Audit Metadata