sglang-model-day0-support
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/collect_public_pr_evidence.pyinvokes the GitHub CLI (gh api) usingsubprocess.runwith a list of arguments. This method avoids shell interpolation and is used strictly for retrieving public metadata. - [EXTERNAL_DOWNLOADS]: The skill retrieves metadata from the public GitHub API. Source integrity is maintained through a configurable repository allowlist that defaults to the official
sgl-project/sglangrepository. - [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from GitHub pull requests to populate its Day-0 bundles. The risk is addressed by the
scripts/validate_day0_bundle.pytool, which scans for sensitive information and malicious patterns, combined with instructions requiring manual verification of all external diffs. - [SAFE]: The inclusion of a dedicated sanitization workflow and a validation script that checks for credentials, private file paths, and PII indicates that the skill is designed with security best practices in mind to prevent accidental data leaks.
Audit Metadata