The Agent Skills Directory

[COMMAND_EXECUTION]: Workflow steps 5 and 6 explicitly instruct the agent to execute the git commit command automatically and with "no confirmation needed," which bypasses the standard safety practice of requiring user approval for shell operations.
[PROMPT_INJECTION]: The skill uses specific directives to override the agent's default behavior regarding autonomous command execution. Additionally, the skill's workflow ingests untrusted data from git diff, which could be used for indirect prompt injection. Ingestion points: Data enters via the git status and git diff commands mentioned in the workflow. Boundary markers: There are no markers or delimiters defined to isolate the diff content from the agent's instructions. Capability inventory: The skill can execute shell commands (git commit). Sanitization: No validation or sanitization of the diff content is specified, allowing content in the repository to potentially influence the agent's behavior.

conventional-commit