The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the dynamic context injection syntax (!command) to execute shell commands at load time. These commands fetch recent git history, GitHub issues, and project status files to provide the agent with real-time context. The commands used (such as git log, gh issue list, and cat) are benign developer tools and do not involve sensitive file access, credential harvesting, or unauthorized network activity.
[INDIRECT_PROMPT_INJECTION]: The skill incorporates external and project-specific data into its prompt, which creates a surface for indirect prompt injection attacks if that data contains malicious instructions.
Ingestion points: Data is ingested from shell command outputs, including git log, gh issue list, and local files in the production/session-state/ and production/milestones/ directories.
Boundary markers: The skill does not implement explicit delimiters or instructions to notify the model that the injected shell output is data that should not be followed as instructions.
Capability inventory: The skill has access to the Write tool for file modifications and the AskUserQuestion tool for interacting with the user.
Sanitization: There is no evidence of sanitization, validation, or escaping of the content retrieved from the shell commands before it is interpolated into the prompt context.
[COMMAND_EXECUTION]: The skill executes local system commands to retrieve project state information. This is standard behavior for a development-focused workflow and aligns with its stated purpose of sprint planning and task management.

game-sprint-plan