skills/liustack/vibemaster/socrates/Gen Agent Trust Hub

socrates

Pass

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious intent or high-risk vulnerabilities were identified. The skill's operations are confined to the local workspace and serve legitimate productivity and privacy goals.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (git check-ignore and echo) to manage local git configurations. This is used to ensure that the .issues/ directory, which contains user decision records, is not accidentally committed to shared repositories.
  • [PROMPT_INJECTION]: The skill provides an interface for processing user-generated input and logging it to the local filesystem, which represents a surface for indirect prompt injection.
  • Ingestion points: User conversation and problem statements ingested via the Phase 1-5 workflows defined in SKILL.md.
  • Boundary markers: Absent; user thoughts are transcribed directly into decision logs without delimiters.
  • Capability inventory: Local file writing to the .issues/ directory and shell command execution for git management.
  • Sanitization: Absent; the skill does not perform custom validation or sanitization of user input before writing it to files, relying on the underlying model's standard safety guidelines.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 25, 2026, 10:20 AM
Security Audit — agent-trust-hub — socrates