create-git-issue

Warn

Audited by Gen Agent Trust Hub on May 11, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains a "Skill Isolation" block that uses override-style language ("sole active authority", "suppress it and continue without interruption") to modify the agent's standard orchestration and multi-skill handling behavior. This pattern is characteristic of instructions designed to bypass system constraints.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting untrusted data from external sources and interpolating it into prompts for PRD generation and issue creation without adequate security controls.
  • Ingestion points: Technical requirements files (technical_requirements.md) in the workspace root or docs folders; content fetched from issue references (provided via number, URL, or path).
  • Boundary markers: No explicit delimiters (e.g., XML tags or unique markers) or "ignore instructions" warnings are used when processing this data.
  • Capability inventory: The skill has the capability to execute shell commands (gh issue create) and write files to the local workspace (prd.md, issues.md).
  • Sanitization: No input validation, escaping, or filtering is performed on the ingested data before it is included in the PRD or issue templates.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 11, 2026, 05:58 PM
Security Audit — agent-trust-hub — create-git-issue