skills/heliohq/ship/refactor/Gen Agent Trust Hub

refactor

Pass

Audited by Gen Agent Trust Hub on Apr 22, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill is well-structured and implements rigorous safety constraints, such as requiring code verification after every change and prohibiting behavioral changes. It includes a human-in-the-loop step via the AskUserQuestion tool for high-risk structural changes.
  • [PROMPT_INJECTION]: Indirect prompt injection surface (Category 8): The skill ingests untrusted data in the form of the user's source code, which is then processed by sub-agents. This is an architectural requirement for the skill's primary purpose. 1. Ingestion points: Source code files and directories read during the Scan phase in SKILL.md. 2. Boundary markers: No specific delimiters are used to isolate code from agent instructions. 3. Capability inventory: The skill and its sub-agents have access to Bash, Read, Write, Edit, and Agent tools. 4. Sanitization: No explicit sanitization or filtering of code comments or content is performed.
  • [EXTERNAL_DOWNLOADS]: The skill's preamble script (preflight.sh) and initialization command (ship-plugin-root) are vendor-owned resources used for environment setup within the heliohq plugin ecosystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 22, 2026, 11:26 AM
Security Audit — agent-trust-hub — refactor