The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by ingesting and acting upon untrusted external data within a game development repository.
Ingestion points: Processes feature-description from user input, reads Game Design Documents (GDD) within the design/ directory, and monitors Git commit history and external issue trackers during sprint planning.
Boundary markers: Absent. There are no instructions to use delimiters or specific safety wrappers when interpolating this external content into the agent's context.
Capability inventory: Access to Bash, Write, Edit, Read, Glob, Grep, and the ability to chain and invoke other platform skills (e.g., /game-brainstorm, /game-start, /game-launch).
Sanitization: Absent. The instructions do not define any validation, escaping, or filtering for the data read from the project environment.
[COMMAND_EXECUTION]: The skill is granted Bash and Write tool access to perform state management and project updates.
Evidence: The skill instructions explicitly command the agent to initialize and update JSON state files in production/session-state/ and update project metadata files like velocity.json using the shell and file-write capabilities.

game-team-orchestrator