The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified. The skill implements high-integrity safety measures for Git operations, such as disabling repository-provided fsmonitor commands during audits to prevent the execution of untrusted code found within a scanned repository.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted data found in Git repositories (such as commit messages, branch names, or file contents). It implements best practices to ensure this data is treated as data and not instructions.
Ingestion points: Commands like git log, git status, and git show in all scripts read user-provided repository metadata.
Boundary markers: The skill explicitly uses the '--' pathspec delimiter and core.quotePath=false in Git commands to prevent path or argument confusion.
Capability inventory: All scripts are restricted to read-only Git operations, creating local backup files in a dedicated user directory (~/.git-backups/), and managing local Git references.
Sanitization: Scripts like git_export_before_drop.sh sanitize stash messages into safe filenames using sed to prevent path traversal or injection.

git-safety-net