skills/dnsqueen/orcinus-skills/waza/Gen Agent Trust Hub

waza

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the 'waza' CLI binary (v0.40.0). The 'ATTRIBUTION.md' file identifies the source as Microsoft and provides a SHA256 checksum for verification. This is a safe external dependency from a trusted vendor.
  • [COMMAND_EXECUTION]: The skill invokes the 'waza' binary for tasks such as running benchmarks ('waza run'), initializing suites ('waza init'), and generating evaluations ('waza generate'). These are standard operations for the tool's purpose.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted 'eval.yaml' and 'SKILL.md' files.
  • Ingestion points: Files are ingested via 'waza run', 'waza generate', and 'waza dev'.
  • Boundary markers: The tool uses structured YAML for evaluation specifications.
  • Capability inventory: The tool can execute shell commands and perform code-based assertions.
  • Sanitization: The skill relies on the 'waza' CLI's internal processing; no explicit sanitization of embedded instructions is described.
  • Note: This is a low-severity finding inherent to the tool's function.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 08:22 PM