secondbrain-db

Warn

Audited by Gen Agent Trust Hub on May 2, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to download and install the 'sbdb' CLI tool from the author's GitHub repository ('github.com/sergio-bershadsky/secondbrain-db').
  • [COMMAND_EXECUTION]: The skill requires shell command execution to perform file management, indexing, and health checks using the 'sbdb' tool.
  • [REMOTE_CODE_EXECUTION]: The skill supports 'virtual fields' in schemas that execute Starlark code to transform document content; although described as sandboxed, this involves dynamic execution of logic defined in project-level configuration files.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of ingesting and processing untrusted markdown data.
  • Ingestion points: Markdown files and YAML records located in the project's documentation and records directories.
  • Boundary markers: No explicit delimiters or 'ignore' instructions are provided to separate document content from agent instructions.
  • Capability inventory: The agent can perform file reads, writes, deletions, and shell command execution via the 'sbdb' tool.
  • Sanitization: While a Starlark sandbox is mentioned for virtual field logic, no sanitization or escaping of the actual document content is described before it is interpolated into the agent's prompt context.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 2, 2026, 08:25 PM