The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because user-provided input ($ARGUMENTS) is directly interpolated into the system prompts of several specialist subagents (Code Explorer, UX Designer, React Developer, Test Automator, and Code Reviewer).
Ingestion points: User-supplied feature descriptions via $ARGUMENTS are used across all phases of the workflow.
Boundary markers: Absent. The skill does not use delimiters or clear instructions to the subagents to ignore potential command-like patterns within the user-provided requirements.
Capability inventory: The skill possesses significant capabilities, including writing/moving files, executing Git commits, and creating Pull Requests via the GitHub CLI.
Sanitization: No sanitization or validation of the user input is performed before it is embedded in the agent prompts.
[COMMAND_EXECUTION]: The orchestrator performs various shell and CLI operations to manage the development lifecycle.
File Operations: Uses mkdir, mv, and rmdir to organize session artifacts and final documentation.
Git Operations: Utilizes git status, git commit, and EnterWorktree (for branch isolation). The commit process includes staging specific files rather than using a wildcard git add ., which is a positive security practice.
GitHub Integration: Executes gh pr create to automate Pull Request generation. This targets a well-known service and is triggered only upon explicit user approval at a checkpoint.

react-feature-development