The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run() in scripts/build_rebase_todo.py and direct shell commands in SKILL.md to interact with the Git CLI. These operations (e.g., git log, git rebase, git filter-branch) are central to the skill's stated purpose of commit rewriting and are implemented using best practices, such as list-based argument passing to avoid shell injection.
[DYNAMIC_EXECUTION]: The skill automates interactive rebasing by setting GIT_SEQUENCE_EDITOR to a cp command that replaces the rebase todo list with a pre-generated file (/tmp/rebase_todo.txt). This allows the agent to modify commit history programmatically based on user-defined mappings.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data, which is analyzed as follows:
Ingestion points: Processes commit_rewrite_mapping.md (user-editable mapping table) and existing commit messages via git log in scripts/build_rebase_todo.py.
Boundary markers: No specific boundary markers are used; data is parsed line-by-line using regex and markdown table structure.
Capability inventory: The skill can execute git commands (rebase, filter-branch), write temporary files, and modify the repository's commit history.
Sanitization: The scripts/build_rebase_todo.py script uses splitlines() when reading git logs, ensuring that only the first line of a commit message is used in the rebase todo file, which prevents multi-line injection into the git rebase sequence. Message mappings are handled via JSON serialization in scripts/msg_filter.py to ensure safe data handling.
[SAFE]: No evidence of data exfiltration, hardcoded credentials, obfuscation, or attempts to bypass safety guidelines was found. The skill operates locally on the user's repository using the provided scripts.

pavlo-commit-rewriter