cookbook-review-pr

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh) to perform read-only operations on the sgl-project/sglang repository, such as viewing pull request details, diffs, and metadata (SKILL.md steps 1-3). These operations are essential to the skill's stated purpose.
  • [COMMAND_EXECUTION]: The skill executes mint validate and mint broken-links to ensure the integrity of the documentation build. These are standard commands for the Mintlify platform used by the SGLang project.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from external pull request content, which represents a surface for indirect prompt injection.
  • Ingestion points: Pull request titles, bodies, author names, and code diffs retrieved via gh commands.
  • Boundary markers: None identified in the instructional steps.
  • Capability inventory: Shell command execution via the gh and mint tools.
  • Sanitization: None identified; the skill relies on a structured checklist for processing the data.
  • [SAFE]: No obfuscation, hardcoded credentials, or persistence mechanisms were detected. The external tools and targeted repositories are well-known and consistent with the vendor's own infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 07:09 AM
Security Audit — agent-trust-hub — cookbook-review-pr