The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of command-line tools including git, gh (GitHub CLI), flutter, and dart. These are used for version control, repository management, and code quality verification (analysis and testing). All commands are standard for the described development workflow.
[DATA_EXFILTRATION]: While the skill involves pushing code to GitHub, it explicitly warns against committing sensitive files like .env, keys, or certificates. Interactions are limited to legitimate repository management via authorized tools.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it reads git diff and git status outputs to generate commit messages and PR descriptions.
Ingestion points: Reads local repository state and file differences.
Boundary markers: Uses task descriptions but lacks strict delimiters for the untrusted diff data.
Capability inventory: Can execute shell commands and interact with the GitHub API.
Sanitization: Relies on the user/agent to review generated messages before execution. The risk is considered low and inherent to the functionality.

git-github