The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is configured to fetch its core engine from an external GitHub repository (https://github.com/tanis90/coc-cardwright.git) which is the author's own infrastructure. This is documented in the ENGINE_REPO.txt file and used by the agent during setup.
[REMOTE_CODE_EXECUTION]: The skill instructions guide the AI agent to clone the remote repository and perform a local installation using 'pip install -e'. This results in the execution of code downloaded from the internet on the host system to provide the character generation functionality.
[COMMAND_EXECUTION]: The skill requires the agent to run multiple shell commands to manage the environment and generate output, specifically 'git' for cloning the repository, 'pip' for package management, and a custom 'coc' CLI tool for data validation and HTML rendering.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes untrusted user input (character descriptions and concepts) which is then passed to the CLI tool and formatted into files.
Ingestion points: User prompts describing character identity, traits, and background (SKILL.md).
Boundary markers: The skill enforces a strict JSON schema for character data and provides detailed field requirements to delimit user content.
Capability inventory: The agent has access to shell execution for 'git', 'pip', and the 'coc' CLI (SKILL.md).
Sanitization: The skill mandates the use of 'coc verify' to validate all generated data against rule-based constraints before rendering, which helps sanitize numerical and structural data.

coc-480buy-card