tsg-flow-orchestrator

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The orchestrator relies on the scripts/tsg-delegate.sh shell script to interact with the environment. This script executes the herdr CLI for agent management, jq for JSON processing, and git for repository state management. These are intended behaviors for a development-focused orchestration tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data sources including task definition files (tasks.md) and output files from sub-agents (result.json, report.md). This creates a vulnerability surface for indirect prompt injection if those files are controlled by a malicious actor. However, the orchestrator uses structured JSON schemas and isolated execution contexts (via Herdr panes) to minimize the risk of instructions in data being executed by the orchestrator itself.
  • Ingestion points: Processes PRD task files and worker result logs from the .tsg-flow/ directory.
  • Boundary markers: Employs explicit delimiters and instruction headers in prompts sent to sub-agents.
  • Capability inventory: Performs file I/O, Git operations, and subprocess execution of the herdr and jq utilities.
  • Sanitization: Uses jq to safely generate and parse JSON payloads and bash's printf %q to escape command arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 05:33 PM