sdl-mcp-symbol-delta-ledger

Warn

Audited by Gen Agent Trust Hub on Jul 11, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The sdl.runtime.execute tool permits the agent to invoke system executables such as npm, node, and python3. While intended for test execution, this provides a mechanism for the agent to run arbitrary scripts or commands within the local environment if manipulated.
  • [EXTERNAL_DOWNLOADS]: The skill's setup instructions require users to download and install the sdl-mcp package and its dependencies from the public NPM registry.
  • [REMOTE_CODE_EXECUTION]: The documentation encourages the use of npx --yes sdl-mcp@latest, which facilitates the direct download and execution of code from a remote registry.
  • [PROMPT_INJECTION]: The skill represents an indirect prompt injection surface as it is designed to ingest and process arbitrary codebase content that could contain hidden instructions designed to manipulate the agent's behavior during analysis.
  • Ingestion points: Codebase indexing and symbol retrieval via sdl.symbol.* and sdl.iris.* tools.
  • Boundary markers: No explicit boundary markers or instruction-isolation logic is defined in the skill documentation.
  • Capability inventory: The skill provides access to raw source code (sdl.iris.rawSource) and the ability to execute commands (sdl.runtime.execute).
  • Sanitization: There are no mentions of sanitization or safety filters applied to the content retrieved from the codebase.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 11, 2026, 12:33 AM
Security Audit — agent-trust-hub — sdl-mcp-symbol-delta-ledger