docs-steward

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill functions as an orchestrator for external markdown formatting tools. As implemented in scripts/docs_steward/process.py, it executes local binaries using subprocess.run with shell=False. This prevents shell injection vulnerabilities. Arguments are passed as an explicit list, and binary resolution is handled via shutil.which against a pre-defined set of common installation paths.\n- [EXTERNAL_DOWNLOADS]: The file scripts/docs_steward/hints.py contains a reference to https://dprint.dev/install.sh within a string literal. This is intended to be displayed to the user as a suggested installation method for the dprint tool. The skill's logic does not perform any network requests or automated execution of this remote script; it is used only for informative reporting.\n- [SAFE]: The skill adheres to the 'no-auto-install' principle and operates locally on the repository files. No evidence of data exfiltration, obfuscation, or prompt injection was found during the analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 10:51 PM
Security Audit — agent-trust-hub — docs-steward