The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to create directories and initialize git repositories. It interpolates user-provided project names directly into shell commands (e.g., mkdir -p ~/projects/{name}), which presents a command injection risk if the user provides input containing shell metacharacters.
[DATA_EXFILTRATION]: Performs a POST request via curl to a local API (localhost:3110) to create project tickets. This operation is restricted to the local environment and uses project metadata rather than sensitive files.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by incorporating user-provided descriptions and tech stack details into the CLAUDE.md file. \n
Ingestion points: User input captured via AskUserQuestion in the interview phase. \n
Boundary markers: Absent; user input is interpolated directly into templates without delimiters. \n
Capability inventory: The skill has Write and Bash capabilities which allow it to create files and execute commands. \n
Sanitization: No input validation or sanitization is performed on the captured strings before they are written to files.

project-kickoff