Skills
skills/team-attention/hoyeon/discuss/Gen Agent Trust Hub

discuss

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines clear boundaries for the agent's role as a thought partner and explicitly forbids prescriptive actions such as generating implementation plans, running git commands, or writing implementation code.
  • [COMMAND_EXECUTION]: The skill utilizes the Task tool to delegate codebase exploration to a specialized 'Explore' sub-agent when the --deep flag is used. This is a standard platform feature for contextual gathering.
  • [DATA_EXFILTRATION]: The Write tool is used to save discussion outcomes to a local hidden directory (.hoyeon/discuss/) only after user confirmation via AskUserQuestion. This represents legitimate local data persistence within the project workspace.
  • [PROMPT_INJECTION]: The skill contains comprehensive behavioral instructions for a Socratic dialogue flow (DIAGNOSE, PROBE, SYNTHESIZE) that guide interaction without attempting to bypass or override system-level safety constraints.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 01:29 PM
Security Audit — agent-trust-hub — discuss