The Agent Skills Directory

[PROMPT_INJECTION]: The framework is susceptible to indirect prompt injection (Category 8) due to its core functionality of fetching and processing data from external platforms such as Jira and Figma. \n
Ingestion points: Data is retrieved from Jira issue descriptions and comments, and Figma design context in sub-skills/spec/SKILL.md. \n
Boundary markers: Although the specification template includes an isolation section (## Input bruto), there are no explicit instructions for the AI to ignore embedded directives within that content during analysis. \n
Capability inventory: The skill has high-privilege tool access, including file system modification, shell execution via git/gh, and sub-agent spawning. \n
Sanitization: No automated filtering or sanitization of external platform data is performed before it enters the agent's context. \n- [COMMAND_EXECUTION]: The skill implements a hooks system (GOD/hooks.md) that executes natural language instructions at various lifecycle steps. This mechanism provides a way for instructions stored in configuration files to influence agent behavior dynamically. \n- [COMMAND_EXECUTION]: Internal Python scripts in the sub-skills/_lib/ directory interact with the shell using subprocess.run to execute git commands based on task metadata stored in local project files. \n- [EXTERNAL_DOWNLOADS]: The skill performs authorized network operations to well-known developer services, including GitHub, Jira, and Figma. These operations are essential for synchronizing the development state and managing project metadata.

god