truto-unified-mappings

Fail

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions for the user to install the Truto CLI by downloading and piping shell scripts from cli.truto.one directly into bash (Linux/macOS) or iex (Windows). These are vendor-provided setup scripts for the required tooling.
  • [EXTERNAL_DOWNLOADS]: The CLI automatically downloads a local ONNX embedding model (all-MiniLM-L6-v2) from a remote source during its first execution to facilitate local documentation search.
  • [PROMPT_INJECTION]: The skill ingests data from external sources, such as API documentation URLs and live responses from connected service accounts, to generate JSONata mapping code via an LLM. This creates a vulnerability to indirect prompt injection if the ingested content contains adversarial instructions.
  • Ingestion points: Untrusted data is retrieved from remote URLs provided via the --source-url flag and from live API samples via the fetch_proxy_sample tool.
  • Boundary markers: The instructions do not specify the use of strict delimiters or explicit warnings to separate external data from the agent's core instructions.
  • Capability inventory: The agent possesses the capability to fetch web content (web_fetch), access connected API accounts, and execute generated logic through a JSONata evaluator (eval_jsonata).
  • Sanitization: The skill documentation does not mention protocols for sanitizing or filtering external API documentation before it is processed by the LLM.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.truto.one/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 21, 2026, 10:13 AM
Security Audit — agent-trust-hub — truto-unified-mappings