luna-executor
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized operations were detected in the skill instructions. The skill utilizes a standard agent orchestration pattern (spawn_agent) for task delegation.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a system where a sub-agent executes commands and processes data based on instructions from the parent agent. Although this creates an attack surface for indirect instructions, the skill includes robust mitigation strategies.
- Ingestion points: The message field passed to the luna_worker via the spawn_agent tool call in SKILL.md.
- Boundary markers: The skill instructs the parent agent to define clear acceptance criteria and verification expectations to bound the sub-agent's task.
- Capability inventory: The delegated luna_worker is expected to perform command or script execution, repository exploration, and build, lint, test runs as described in SKILL.md.
- Sanitization: The workflow mandates that the parent agent inspect the actual output and workspace diff and run or confirm risk-proportionate checks before accepting the result.
Audit Metadata