dispatching-parallel-agents

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
  • [NO_CODE]: The skill consists exclusively of Markdown documentation and does not include any executable code, scripts, or configuration files that trigger automated actions.\n- [PROMPT_INJECTION]: The skill describes a pattern for creating sub-agent tasks using external data like test failure logs and source code contents, which introduces an indirect prompt injection surface.\n
  • Ingestion points: Test failure details and file contents (e.g., 'src/agents/agent-tool-abort.test.ts') are used to define sub-tasks in 'references/examples.md'.\n
  • Boundary markers: The provided examples do not use explicit delimiters or instructions to isolate potentially untrusted task data from instructions.\n
  • Capability inventory: The orchestration pattern utilizes commands to spawn sub-tasks which are described as having the capability to read from and write to the local filesystem.\n
  • Sanitization: There is no demonstration or instruction regarding the sanitization or escaping of external data before it is interpolated into task prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 04:41 PM