The Agent Skills Directory

[Prompt Injection] (LOW): The skill documentation includes examples of interpolating untrusted user data into shell commands, which is a classic surface for indirect prompt injection.
Ingestion points: The variables {user_intent} and {task_description} in SKILL.md are used to populate command-line arguments for the reasoning-index tool.
Boundary markers: There are no boundary markers or explicit instructions to the agent to treat the interpolated content as untrusted data or to sanitize it for shell metacharacters.
Capability inventory: The skill is designed to trigger a variety of actions including code implementation, git operations (commit, push, merge), and system architecture tasks through the reasoning-index and sc:* command set.
Sanitization: The provided documentation shows no evidence of input validation, escaping, or schema enforcement before the user-provided strings are passed to the CLI tool.

commands-router