skills/letta-ai/skills/memfs-search/Gen Agent Trust Hub

memfs-search

Pass

Audited by Gen Agent Trust Hub on May 24, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads approximately 2GB of local embedding models from Hugging Face during the setup process. Hugging Face is a well-known and trusted platform for model distribution.
  • [COMMAND_EXECUTION]: The provided bash script wraps the qmd command-line tool to perform indexing and searching. It uses standard argument passing and does not appear to contain command injection vulnerabilities.
  • [DATA_EXFILTRATION]: While the skill accesses the agent's memory directory ($MEMORY_DIR), this is the primary purpose of the skill. The search operations and embedding generations are performed locally, with no evidence of data being transmitted to external servers.
  • [SAFE]: The skill uses local resources for AI operations (GGUF models via QMD) which enhances privacy by avoiding external API calls for embeddings.
Audit Metadata
Risk Level
SAFE
Analyzed
May 24, 2026, 05:30 AM
Security Audit — agent-trust-hub — memfs-search