vibe-coding-orchestrator

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOW
Full Analysis

{ "verdict": "LOW", "summary": "The skill is an orchestration framework for development tasks like linting, testing, and security. It defines complex workflows for an AI agent but does not contain active malicious code or unauthorized data exfiltration. The primary risks are the indirect prompt injection surface and character encoding corruption in one of the metadata files.", "categories": [ "PROMPT_INJECTION", "COMMAND_EXECUTION" ], "detailed_analysis": "- [Indirect Prompt Injection] (LOW): The skill uses automatic activation rules (workflow/auto-activation.md) that trigger specific 'AI Skills' and workflows based on file names and content patterns. This creates a vulnerability surface where an attacker providing malicious files to the agent could influence the agent's internal reasoning and selection of tools.\n

  • Evidence:\n
  • Ingestion points: The skill uses Read, Glob, and Grep to scan the local codebase.\n
  • Boundary markers: None identified in the orchestration logic to distinguish between instructions and code data.\n
  • Capability inventory: High (uses Task tool and orchestrates command execution).\n
  • Sanitization: Not present; the logic relies on regex matching of raw content.\n- [Obfuscation] (LOW): The file core/skill-registry.md contains significant character encoding corruption, resulting in unreadable Unicode replacement characters in the description fields (e.g., ] ). While likely a non-malicious technical error, it constitutes a form of unintentional obfuscation that prevents the user from understanding the full registry.\n- [Command Execution] (INFO): The skill is designed to prompt the agent to execute standard development commands (e.g., npm run lint, npm run test, vibe check). While these are legitimate developer tools, users should be aware that the 'Vibe' orchestration logic will actively drive the agent to run these processes via the Task tool.", "python_packages": [], "node_packages": [ "vitest", "eslint", "prettier", "husky", "lint-staged" ], "remote_code_patterns": [] }
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 08:42 AM