GitHub Agentic Workflows Operations

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill documents the use of the bash tool for executing shell commands within the workflow environment for tasks like compilation and deployment.
  • [EXTERNAL_DOWNLOADS]: The skill provides patterns for downloading and integrating remote components, including MCP servers from HashiCorp and Microsoft, as well as reusable agents and skills from external GitHub repositories.
  • [PROMPT_INJECTION]: Implementation patterns for IssueOps and Code Analysis workflows involve processing untrusted data from GitHub issues, comments, and pull requests, creating a surface for indirect prompt injection.
  • Ingestion points: GitHub event triggers such as issues, pull_request, and issue_comment described in Skill 7.
  • Boundary markers: Not explicitly defined in the provided workflow templates to segregate untrusted data from instructions.
  • Capability inventory: Includes powerful tools such as bash (command execution) and edit (file system modification).
  • Sanitization: No specific methods for sanitizing or validating ingested user content are mentioned in the guide's examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 08:06 PM