The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Git commands via a Python script using the subprocess.run method. The script correctly uses list-based arguments without the shell=True flag, which prevents shell injection attacks.\n- [COMMAND_EXECUTION]: Input validation is strictly enforced in scripts/resolve_branch.py. The slugify function uses Unicode normalization and regex filtering to ensure user-provided descriptions are safe for use as branch names.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined task descriptions to generate branch names, representing a potential indirect injection surface.\n
Ingestion points: User-provided --slug and --type arguments in scripts/resolve_branch.py.\n
Boundary markers: None present in the interpolation logic.\n
Capability inventory: Local Git command execution via subprocess.\n
Sanitization: Robust character filtering in the slugify function and regex validation using SAFE_BRANCH_RE.

harness-worktree