github-issue-repro-and-fix

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection from external data sources.
  • Ingestion points: Reads untrusted content from GitHub issue reports, pull requests, and comments using the gh CLI (SKILL.md).
  • Boundary markers: Lacks instructions or markers to treat external content as untrusted or to ignore embedded instructions.
  • Capability inventory: Executes potentially user-influenced code through pnpm test, pnpm typecheck, and pnpm lint (SKILL.md).
  • Sanitization: No evidence of sanitization or filtering of external issue content before it is processed by the agent.
  • [COMMAND_EXECUTION]: Employs standard developer tools including gh, rg, and pnpm to analyze the codebase and verify fixes. These commands are aligned with the skill's intended purpose of repository maintenance.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 01:08 PM
Security Audit — agent-trust-hub — github-issue-repro-and-fix