orchestrator

Pass

Audited by Gen Agent Trust Hub on May 3, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it interpolates user-supplied task descriptions into agent prompts via the $ARGUMENTS variable. Evidence: 1. Ingestion points: SKILL.md (argument-hint and prompt). 2. Boundary markers: Absent. 3. Capability inventory: Bash, Task, Edit, Write, and LSP tools. 4. Sanitization: Absent. This allows potentially malicious instructions to influence the orchestration workflow.
  • [COMMAND_EXECUTION]: The skill relies heavily on the Bash tool to execute local utility scripts such as .claude/scripts/glm5-teammate.sh and .claude/lib/action-report-lib.sh, as well as CLI commands like ralph.
  • [DYNAMIC_EXECUTION]: The skill uses the Task tool to dynamically spawn and coordinate subagents (e.g., ralph-coder, ralph-reviewer) based on a plan it generates and stores in .claude/orchestrator-analysis.md.
Audit Metadata
Risk Level
SAFE
Analyzed
May 3, 2026, 11:06 AM
Security Audit — agent-trust-hub — orchestrator