arena
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by accepting external task data and passing it to a secondary workflow.
- Ingestion points: The skill instructions specify processing the 'caller's task, candidate count, runner preferences and scope' (SKILL.md).
- Boundary markers: No specific delimiters or warnings are provided to differentiate untrusted task content from instructions.
- Capability inventory: The skill identifies a capability to execute a sibling markdown file (
../agi-mode/playbooks/arena.md) and interact with an existing Hillclimb record. - Sanitization: No data validation or escaping mechanisms are mentioned for the ingested caller tasks.
Audit Metadata