orchestrator
Pass
Audited by Gen Agent Trust Hub on Aug 5, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill focuses on establishing a disciplined orchestration workflow. It includes security-positive instructions, such as requiring direct evidence of task completion and warning against practices that mask command exit codes (e.g., piping through grep or tail). No suspicious behaviors were identified.\n- [PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by design as it coordinates and processes output from other agents.\n
- Ingestion points: The Reviewer / checker policy in SKILL.md processes the WORKER RETURN, which is the full output from a subagent.\n
- Boundary markers: The Review Prompt Template uses labels like WORKER RETURN:, but lacks specific instructions to ignore malicious directives embedded in that return.\n
- Capability inventory: The orchestrator agent is authorized to execute commands (rerun key tests) and modify files (mechanical-fix carve-out) based on its evaluation of the worker's output.\n
- Sanitization: The instructions do not specify any sanitization or validation logic for the ingested worker output.
Audit Metadata