create-issues

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it extracts instructions and task descriptions from untrusted codebase data.
  • Ingestion points: The agent reads git diff, git status, git log, and various source files mentioned in the conversation to identify potential issues and action items.
  • Boundary markers: The workflow includes a mandatory "Review gate" (Step 3) that requires the agent to present all candidate issues to the user for explicit approval and editing before proceeding to execution.
  • Capability inventory: The skill has the ability to write local files to docs/planning/issues/, use the gh CLI for issue and project management, and execute GraphQL mutations via gh api to manage issue dependencies.
  • Sanitization: The instructions include specific implementation guidance for constructing bash strings (using printf or $'...') to ensure that issue bodies containing special characters or newlines do not cause command injection or formatting errors when passed to CLI tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 02:09 PM
Security Audit — agent-trust-hub — create-issues