Skills
skills/broomva/agent-control-metalayer-skill/control-metalayer-loop/Gen Agent Trust Hub

control-metalayer-loop

Pass

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The template scripts (e.g., smoke.sh, check.sh, test.sh) utilize the eval command to execute arbitrary instructions provided via environment variables such as CONTROL_CHECK_CMD, which allows for dynamic command overrides.
  • [COMMAND_EXECUTION]: The skill automates the configuration of git hooks by setting core.hooksPath to a local .githooks directory, enabling the execution of scripts during git operations like commits and pushes.
  • [COMMAND_EXECUTION]: The bootstrap process dynamically modifies the repository's Makefile to include Makefile.control, which establishes an integrated execution surface for the agent's control targets.
  • [PROMPT_INJECTION]: The skill implements a surface for indirect prompt injection by processing instructions from .control/commands.yaml. Ingestion points: .control/commands.yaml. Boundary markers: None identified. Capability inventory: eval in shell templates and subprocess.run in control_wizard.py. Sanitization: No sanitization or validation logic for command interpolation was found.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 17, 2026, 07:59 AM