The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the user to clone source code from an external, untrusted GitHub repository (https://github.com/tianxingleo/ACGTI.git) that is not associated with a trusted organization or the skill author's known vendor infrastructure.
[COMMAND_EXECUTION]: The installation instructions involve running npm install, npm run dev, and npm run build on the external project. These commands can execute arbitrary shell scripts (via preinstall, postinstall, or build hooks) on the host system, which could be used for unauthorized activities if the repository contains malicious configuration.
[INDIRECT_PROMPT_INJECTION]: The skill architecture is designed to ingest and process data from local JSON files (questions.json, characters.json) that are provided by the external repository. This creates a surface for indirect prompt injection where an attacker controlling the repository could embed instructions in the data to influence the agent's behavior.
Ingestion points: src/data/questions.json, src/data/characters.json, src/data/archetypes.json, and src/data/characterVisuals.json.
Boundary markers: None are specified to prevent the agent from interpreting instructions found within these data files.
Capability inventory: The skill environment has capabilities for file system access, network operations for deployment, and shell command execution.
Sanitization: No sanitization or validation logic for the external JSON content is described, allowing potentially malicious instructions to be processed by the quiz engine or the agent.

acgti-anime-persona-quiz