mcore-create-issue
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
ghcommand-line interface to interact with GitHub repositories. It performs operations such as viewing workflow runs (gh run view), fetching logs and commit metadata through the GitHub API (gh api), and automating the creation of repository issues (gh issue create). All operations are scoped to theNVIDIA/Megatron-LMrepository. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external, untrusted data from GitHub Actions logs and pull request metadata to generate issue content.
- Ingestion points: Untrusted CI logs are retrieved via
gh api repos/NVIDIA/Megatron-LM/actions/jobs/<job_id>/logsinSKILL.md. - Boundary markers: The instructions mandate the use of markdown code blocks to wrap log outputs, which serves as a delimiter between instructions and processed data.
- Capability inventory: The skill has the capability to write to the repository by creating new issues with
gh issue createinSKILL.md. - Sanitization: The workflow includes specific constraints, such as limiting the error snippet to 30 lines and requiring verbatim quoting, to mitigate the risk of malicious payload execution through the processed data.
Audit Metadata