dispatching-parallel-agents
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely documentation-based, providing patterns for task delegation and coordination. It does not include scripts, binaries, or network-enabled commands.
- [INDIRECT_PROMPT_INJECTION]: The skill outlines a workflow for processing external data (e.g., test failures and bug reports). It explicitly recommends context isolation ("isolated context", "precisely crafting their instructions") and strict constraints as a core principle. While the agent manages tasks derived from external sources, the skill promotes defensive boundary practices rather than introducing vulnerabilities.
- Ingestion points: Task descriptions and failure logs are ingested into sub-agent prompts as described in
SKILL.md. - Boundary markers: The pattern emphasizes "precisely crafting their instructions" and ensuring agents "never inherit your session's context or history."
- Capability inventory: The skill uses
TaskCreate,TaskList, andTaskUpdatefor workflow orchestration. - Sanitization: The methodology relies on scoping and isolation to mitigate the impact of potentially malicious instructions in task data.
Audit Metadata