arena
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists entirely of markdown instructions defining a logical workflow for an AI agent. It does not contain scripts, binaries, or automated command execution sequences.
- [DATA_EXPOSURE]: No hardcoded credentials or sensitive file access patterns were detected. The skill specifically recommends using isolated temporary directories (e.g.,
arena-<slug>/candidate-<n>/) for file-backed artifacts, which is a security best practice for managing transient data. - [INDIRECT_PROMPT_INJECTION]: The skill manages the risk of indirect prompt injection by design. It instructs the agent to process outputs from multiple subagents (untrusted data) but mitigates risk through isolation techniques, such as using separate model families for judging and enforcing strict rubrics.
- Ingestion points: Candidate outputs generated in Phase B are read by the parent agent and judge subagent in Phase C and D.
- Boundary markers: The skill mandates isolation of outputs via labeled inline responses or separate filesystem paths to prevent context contamination.
- Capability inventory: The workflow involves file system writes and subagent spawning, but these are constrained to the agent's authorized platform tools.
- Sanitization: Phase F (Verify) requires the synthesized result to be scrutinized and proven correct before completion.
Audit Metadata