orchestrate
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill orchestrates subagents and background processes based on the 'agreed work' from the conversation history, creating a surface for indirect prompt injection.
- Ingestion points: The skill uses 'the work agreed in this conversation' as the primary source for task delegation (SKILL.md).
- Boundary markers: There are no instructions to use delimiters, tags, or 'ignore embedded instructions' warnings when the agent translates conversational context into subagent tasks.
- Capability inventory: The skill spawns and manages subagents (Opus 5.5 by default), executes background shell commands (build, lint, test), and performs verification of code diffs.
- Sanitization: No sanitization or validation mechanisms are specified for the conversational input before it is interpolated into subagent instructions.
Audit Metadata