execute-automated-fix

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via the issue context variable, which contains potentially untrusted data such as the issue description and reproduction steps.
  • Ingestion points: The issue object (id, title, description, reproduction steps) used in the 'Understand the issue' and 'Reproduce the issue' steps of the workflow.
  • Boundary markers: Absent. There are no delimiters or instructions to ignore embedded commands within the issue data.
  • Capability inventory: The agent is instructed to modify the project filesystem, add instrumentation, run code to capture logs, and verify fixes by executing reproduction steps.
  • Sanitization: Absent. The skill does not perform validation or filtering of the content provided in the issue context variable before processing it.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 04:49 AM