init-project

Warn

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads project templates from repositories within the humblebeeai organization on GitHub.\n- [COMMAND_EXECUTION]: The skill executes multiple system commands to manage the environment, including gh, git, rsync, chmod, and pre-commit.\n- [REMOTE_CODE_EXECUTION]: The skill executes pre-commit run, which initiates the execution of hooks defined within the downloaded external templates. It also grants execution permissions to shell scripts found in those templates using chmod +x.\n- [PROMPT_INJECTION]: The skill ingests data from external GitHub templates and performs file-writing and adaptation without boundary markers or content sanitization, presenting a surface for indirect prompt injection.\n
  • Ingestion points: SKILL.md (cloning and reading template files)\n
  • Boundary markers: None provided to distinguish template content from agent instructions\n
  • Capability inventory: File system access, network operations (gh, git), and command execution (chmod, pre-commit)\n
  • Sanitization: No sanitization or validation of the template content before interpolation
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 6, 2026, 05:46 AM
Security Audit — agent-trust-hub — init-project