The Agent Skills Directory

[SAFE]: The skill implements a valid professional workflow for validating game designs. It operates entirely within the local repository's directory structure (e.g., design/, docs/, Assets/Prototypes/) and uses standard developer tools for reading, writing, and testing code. No patterns indicative of credential harvesting, unauthorized network communication, or malicious remote execution were detected.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted content from the repository's design and architecture documentation. This identifies a potential surface for indirect prompt injection if project documentation contains instructions intended to manipulate the agent's behavior during the slice implementation.\n
Ingestion points: Reads from project documentation (design/gdd/, docs/architecture/, CLAUDE.md) and configuration files (production/review-mode.txt).\n
Boundary markers: The skill does not employ explicit delimiters or instruction-shielding prompts when interpolating data from these files into the agent context.\n
Capability inventory: The skill uses Read, Write, Edit, Bash, and Task tools, which allow the agent to modify the file system and execute shell commands within the project environment.\n
Sanitization: No sanitization or validation of the content from ingested documents is performed prior to processing.

vertical-slice