skills/alpoxdev/hypercore/hyper-skill/Gen Agent Trust Hub

hyper-skill

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill bundles a bash script (scripts/validate-hyper-skill.sh) intended for structural and content validation of other agent skills. The script is written using standard POSIX utilities (awk, sed, grep, find) and jq for JSON processing, following portability guidelines.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and refactor existing skill folders, which constitutes an ingestion surface for untrusted data. However, the skill's own rules (e.g., rules/context-and-harness-alignment.md) provide explicit instructions for the agent to treat retrieved material as evidence rather than authority and include validation cases specifically for rejecting prompt injections in external data.
  • [EXTERNAL_DOWNLOADS]: The skill references several external documentation sources for the Agent Skills standard and vendor-specific implementations (OpenAI, Anthropic, Vercel). These are well-known technology organizations and their official documentation domains are used for reference and grounding.
  • [DYNAMIC_CONTEXT_INJECTION]: The skill documentation describes the !command syntax supported by certain agent platforms for dynamic context injection. While it explains this capability for educational and authoring purposes, it does not utilize it to execute unauthorized commands at load time.
  • [SAFE]: The skill incorporates a robust "Artifact Audit" process in its rules, specifically instructing users and agents to check for code execution, instruction manipulation, network access, and hardcoded credentials in any skill before distribution. This proactive security posture is a core feature of the tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 07:40 PM