The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess.run with list-based arguments to execute local git commands for analysis and commit creation. This is a standard and safe method for CLI integration.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted data from git diff output and SKILL.md frontmatter to generate commit messages. This data is interpolated into the agent's context. While the skill uses regex and templates to sanitize and structure this input, it remains an attack surface where malicious code in a diff could attempt to influence the agent's subsequent actions.
Ingestion points: scripts/generate_commit_message.py and scripts/categorize_changes.py ingest output from git diff and file contents.
Boundary markers: Absent. The scripts do not explicitly wrap the generated messages in delimiters to warn the agent about embedded instructions.
Capability inventory: The skill executes git commands (diff, commit, write-tree) and documents the use of clawhub and skillhub for publishing.
Sanitization: The analyze_diff_content function in scripts/generate_commit_message.py uses regex patterns to extract specific headers, function names, and keywords rather than passing raw diff content.
[EXTERNAL_DOWNLOADS]: The documentation and code prompt the user to install the PyYAML library from the official Python package index (PyPI) if it is missing.

git-batch-commit