execute-work-package

Warn

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Background Process Management. The scripts/start-l4l-oci.sh script initiates a background service using python -m l4l_oci via nohup and performs health checks using inline Python code. This acts as a local service manager for the execution environment.
  • [COMMAND_EXECUTION]: Subagent Verify Capability. The execution protocol allows subagents to run arbitrary shell commands defined as the verify_command to validate changes. This grants significant privilege to subagents to execute logic within the host environment.
  • [EXTERNAL_DOWNLOADS]: Dependency on Local Environment. The skill relies on a local repository at ~/git/l4l-oci and the fastmcp library. Executing code from a local path that may be modified by other processes or skills introduces a dependency risk.
  • [PROMPT_INJECTION]: Indirect Injection Surface. The subagent ingests content from project documentation and plans to perform implementation actions. Without strict boundaries or sanitization, instructions embedded in these files could influence the subagent's implementation or verification steps.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 20, 2026, 08:08 PM
Security Audit — agent-trust-hub — execute-work-package