The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/submit_pr.py contains a branch_policy function that uses exec(compile(...)) to dynamically load and execute a Python script from a relative path pointing to a sibling skill directory (../git-commit/scripts/branch_name.py). Loading executable content from computed paths is a sensitive capability.
[COMMAND_EXECUTION]: The skill makes extensive use of subprocess.run to call git and gh (GitHub CLI) for core repository operations, such as pushing branches and creating Pull Requests. This behavior is consistent with the skill's primary stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository into Pull Request bodies. (1) Ingestion points: Commit subjects, messages, and diff details are read via git log and git diff. (2) Boundary markers: The ingested content is delimited within a fixed Markdown template (assets/pull_request_template.md). (3) Capability inventory: The skill has the ability to push to remote repositories and create Pull Requests on GitHub. (4) Sanitization: The skill validates PR titles against Angular conventions and scans for auto-closing keywords, but does not sanitize the natural language content extracted from repository logs against potential prompt injection attacks.

submit-pr