The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts in resources/implementation-playbook.md use subprocess.run to execute Git commands via string interpolation. Specifically, the _get_changed_files and _get_change_stats methods construct commands using f"git diff ... {base_branch}". If the base_branch variable is influenced by untrusted input, it could allow for Git argument injection (e.g., using flags like --ext-diff to execute arbitrary binaries), even though shell=True is not used.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (git diffs and commit messages) which could be controlled by an attacker in a collaborative environment.
Ingestion points: Git diff output and commit history processed in PRAnalyzer.analyze_changes and generate_pr_description within resources/implementation-playbook.md.
Boundary markers: None. The skill does not use delimiters or instructions to prevent the agent from following directives found inside the diffs or commits.
Capability inventory: The skill has the ability to execute shell commands via subprocess.run and generate content based on analyzed data.
Sanitization: No filtering or sanitization of the content extracted from Git is performed before it is presented to the agent for analysis.

git-pr-workflows-pr-enhance