continue-in-background

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands to collect local context (git branch, status, logs) and uses the claude --bg CLI command to start background sessions.
  • [PROMPT_INJECTION]: Includes a 'Hard gate' requiring explicit user request before launching background agents, which prevents the model from self-electing potentially expensive or unauthorized background execution.
  • [COMMAND_EXECUTION]: Implements input sanitization on the topic argument, restricting it to a kebab-case alphanumeric pattern ([a-z0-9-]) to prevent the injection of malicious flags or shell meta-characters into the launch command.
  • [PROMPT_INJECTION]: Uses a secure execution pattern for untrusted session text: instead of inlining the resume prompt in the command string, the skill writes it to a temporary file and uses shell indirection (cat "<prompt-file>") to prevent crafted prompts from escaping shell quoting.
  • [DATA_EXPOSURE]: Accesses local project metadata (git history, branch names) to provide context for the handoff save-point. The data handling is restricted to the local filesystem and the new session context.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:55 PM
Security Audit — agent-trust-hub — continue-in-background