The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes shell commands via the gh and icafe-cli tools to create issues on external platforms. This creates a potential command injection surface if the PRD or SPEC files contain shell metacharacters that are not escaped by the agent during command construction. It also performs local file system operations using mkdir -p and file writing.
[PROMPT_INJECTION]: The skill processes external, untrusted PRD and SPEC documents to generate implementation tickets, which constitutes an indirect prompt injection vulnerability. A malicious document could contain instructions intended to override the agent's logic. 1. Ingestion points: PRD and SPEC files located in the tasks/ directory or user-specified paths (SKILL.md). 2. Boundary markers: The instructions lack delimiters or warnings to the agent to treat external file content as data only. 3. Capability inventory: Accesses the shell to run gh and icafe-cli and the file system to write markdown files. 4. Sanitization: No sanitization or validation steps are defined for the content extracted from the requirement documents.

to-issues