memory-ingest

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (mi cortex ingest, mi cortex sync-vault, and mi cortex status) to process and sync local files.
  • [DATA_EXFILTRATION]: The skill's primary function is to read local content, including potentially sensitive Obsidian vaults or personal documentation, and upload it to the vendor's remote service ('Mitosis memory'). While intended, this constitutes a data transmission of local file content.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local Markdown files and directories, creating a vulnerability surface where malicious instructions embedded in those files could influence future agent actions upon retrieval.
  • Ingestion points: The <paths...> and <dir> arguments used in the mi cortex ingest and mi cortex sync-vault commands (SKILL.md).
  • Boundary markers: None specified; there are no instructions for the agent to use delimiters or ignore embedded prompts within the ingested files.
  • Capability inventory: The skill utilizes command execution for file reading and network ingestion through the mi CLI (SKILL.md).
  • Sanitization: No sanitization, escaping, or validation of the file content is described before the ingestion process.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 06:18 AM
Security Audit — agent-trust-hub — memory-ingest