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

hyper-loop

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill incorporates clear defensive instructions that require the agent to treat external content, including logs and evidence files, strictly as data. It explicitly forbids interpreting such content as instructions, effectively mitigating potential indirect prompt injection attacks.
  • [COMMAND_EXECUTION]: The skill relies on a local bash script and standard utilities like git and jq to manage and verify its state. It does not perform any remote code downloads or execute unverified external scripts.
  • [DATA_EXFILTRATION]: The instructions explicitly define a security boundary by excluding remote delivery operations (e.g., git push, merge, or release) from the skill's scope, ensuring that all work remains local to the user's environment.
  • [DYNAMIC_EXECUTION]: The skill uses cryptographic hashing (SHA-256) and git-based worktree fingerprints to verify that task evidence is current and has not been invalidated by subsequent code changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 07:40 PM