parallel-agents
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process and reconcile data from external, untrusted sources which may contain malicious instructions designed to influence child agents.
- Ingestion points: The skill ingests data from logs, web pages, source files, issues, API responses, and generated artifacts (as described in the 'Security Model' and 'integration-and-failures.md').
- Boundary markers: The skill enforces 'context isolation' and 'focused briefs,' ensuring child agents do not inherit the parent transcript, which limits the cross-contamination of potentially malicious instructions.
- Capability inventory: The skill dispatches agents that perform file operations, code analysis, and shell command execution within isolated workspaces managed by 'worktree-isolation'.
- Sanitization: The instructions explicitly mandate that 'instruction-shaped content' found in external data must be treated as data and never used to expand scope, authorize mutations, or change orchestration behavior.
Audit Metadata