skills/lev-os/agents/ultraqa/Gen Agent Trust Hub

ultraqa

Pass

Audited by Gen Agent Trust Hub on Jul 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes arbitrary shell commands defined in the repository for building, testing, linting, and type-checking (e.g., npm test, make).
  • [COMMAND_EXECUTION]: The skill generates temporary scripts, test fixtures, and harnesses at runtime and executes them to verify behavioral scenarios.
  • [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection as it ingests untrusted data from the repository (code, test outputs, and state files) and uses it to drive automated diagnosis and code fixing via the architect and executor prompts.
  • Ingestion points: Reads repository context, command outputs, and .omx/state files.
  • Boundary markers: Explicitly mentions safety bounds in the instructions, though specific delimiters for untrusted data interpolation are not detailed.
  • Capability inventory: Can execute shell commands, write files to the repository, and run generated scripts.
  • Sanitization: Includes internal checks to avoid destructive commands and secret exfiltration.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 16, 2026, 04:48 PM
Security Audit — agent-trust-hub — ultraqa