The Agent Skills Directory

[COMMAND_EXECUTION]: The implementation playbook contains Python code that constructs shell commands using f-strings and executes them via subprocess.run. This pattern is vulnerable to argument injection if parameters like branch names are provided by a user.
Evidence: The _get_changed_files and _get_change_stats methods in sub-skills/implementation-playbook.md build git diff commands using the base_branch variable.
[PROMPT_INJECTION]: The skill is designed to process external, untrusted data (commit messages and code diffs) to generate documentation but lacks mechanisms to prevent indirect prompt injection from that data.
Ingestion points: sub-skills/implementation-playbook.md (processing git diffs and commit history via PRAnalyzer).
Boundary markers: Absent; the skill does not use delimiters or instructions to ignore embedded commands in the processed data.
Capability inventory: The skill performs subprocess execution (git commands) and generates structured text output.
Sanitization: There is no evidence of sanitization, validation, or escaping of the data ingested from the git repository before it is used to generate the final response.

git-pr-workflows-pr-enhance