mcore-split-pr

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands using gh (GitHub CLI) and git to analyze PRs and manage branches. These operations are scoped to the NVIDIA/Megatron-LM repository and user forks, minimizing the risk of unauthorized access or broader system impact.
  • [PROMPT_INJECTION]: The skill processes external, untrusted content from GitHub pull requests (titles and bodies) which presents a surface for indirect prompt injection. However, the risk is mitigated by the workflow's requirement for human review and the lack of high-risk capabilities like sensitive file access or direct network exfiltration.
  • Ingestion points: Data retrieved from gh pr view and gh pr diff in SKILL.md.
  • Boundary markers: None explicitly defined to isolate PR content from instructions.
  • Capability inventory: PR creation, git branch management, and file modification through standard CLI tools.
  • Sanitization: No automated sanitization of pull request text is implemented; the agent relies on its internal safety filters and user oversight.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 04:35 PM
Security Audit — agent-trust-hub — mcore-split-pr