gstack-design-consultation

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted data from local files and web search results, creating a potential attack surface.\n
  • Ingestion points: The agent reads README.md and package.json in SKILL.md Phase 0, and retrieves external information via WebSearch in Phase 2.\n
  • Boundary markers: There are no explicit delimiters or system instructions provided to ensure the agent ignores malicious prompts that might be embedded in the project files or search results.\n
  • Capability inventory: The agent has the ability to execute shell commands via Bash and create files using the Write tool.\n
  • Sanitization: The skill does not implement validation or escaping for the content it ingests from external or local sources.\n- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute system commands such as ls, cat, and head for product context gathering.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 05:57 PM
Security Audit — agent-trust-hub — gstack-design-consultation