The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its processing of external tweet data.
Ingestion points: In tweet-pipeline.md, the skill fetches and parses third-party tweet content to derive game concepts.
Boundary markers: The skill includes explicit safety instructions in tweet-pipeline.md directing the agent to ignore any code or instructions within tweets and use the content only for creative inspiration.
Capability inventory: The skill has extensive capabilities including shell command execution (npm, node, bash), file system modification, and network access via curl and WebFetch.
Sanitization: Mitigation is based on instructional boundary markers provided to the agent.
[COMMAND_EXECUTION]: The skill performs several automated system and environment management tasks.
Executes build and test pipelines using npm run build and npm test.
Manages local credentials for the here.now hosting service by writing to ~/.herenow/credentials and setting restrictive file permissions.
Instructions include modifying the agent's global configuration by adding an MCP tool via claude mcp add playwright.
[EXTERNAL_DOWNLOADS]: The skill integrates various external dependencies and remote resources.
Downloads game assets and 3D models from services including Meshy AI, World Labs, and Sketchfab.
Injects the Play.fun Browser SDK directly into the game's HTML from https://sdk.play.fun/latest.
These downloads originate from well-known technology platforms and service providers.

viral-game