git-worktree-isolation

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses Python's subprocess.run and shell scripts to execute Git commands for repository state inspection. These operations are restricted to standard, read-only metadata queries (e.g., git rev-parse, git status) and utilize safe argument passing (list-based) rather than shell interpolation.
  • [INDIRECT_PROMPT_INJECTION]: The skill contains a dedicated security model that instructs the agent to treat all repository content, logs, and generated files as untrusted data, specifically warning against following instructions found within those sources.
  • [DATA_EXFILTRATION]: Explicit safety invariants prohibit the agent from copying sensitive configuration files (such as .env, tokens, or keys) between workspaces and mandate the use of project-approved secret management mechanisms.
  • [DYNAMIC_EXECUTION]: The skill generates environment metadata at runtime via localized scripts; however, these scripts are built-in, deterministic, and used exclusively for workspace discovery rather than executing arbitrary remote content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 09:12 AM
Security Audit — agent-trust-hub — git-worktree-isolation