The Agent Skills Directory

[SAFE]: The skill instructions and reference files contain no evidence of prompt injection, malicious command execution, or unauthorized data exfiltration. The methodology explicitly prohibits copying protected intellectual property and focuses on original gameplay design.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (SOURCE_BIBLE.md and PRODUCT_BRIEF.md), creating a surface for indirect prompt injection. • Ingestion points: The agent reads the entire contents of user-provided markdown files for analysis. • Boundary markers: The instructions do not specify the use of clear delimiters (e.g., XML tags) to separate source text from analysis instructions. • Capability inventory: The skill is restricted to reading local markdown files and writing output to concepts/CONCEPT.md. It does not have access to network tools, shell execution, or sensitive environment variables. • Sanitization: No input sanitization or filtering of the source text is performed. • Mitigation: Although the risk is negligible due to the absence of dangerous capabilities, implementation of boundary markers for the source text is recommended.

game-concept