The Agent Skills Directory

[SAFE]: No malicious code, prompt injections, or obfuscated patterns were detected. The skill is an educational and architectural guide for organizing agentic workflows.
[DATA_EXPOSURE]: The skill defines a local filesystem-based state management system using the data/ directory. It explicitly identifies the use of hardcoded API keys as an anti-pattern and recommends using environment variables for sensitive data.
[PROMPT_INJECTION]: The architecture includes surfaces for potential indirect prompt injection as it processes project files and agent definitions provided in the environment.
Ingestion points: Files located in data/, agents/, and the CLAUDE.md kernel.
Boundary markers: None explicitly defined in the provided markdown templates.
Capability inventory: Full filesystem access within the project root, Git operations, test runner access, and execution of shell scripts via the scripts/ layer.
Sanitization: No specific content sanitization or filtering logic is described for the data being ingested.
[COMMAND_EXECUTION]: The skill documents how to set up scheduled automation using system-standard tools such as macOS LaunchAgents and Linux systemd timers. These configurations are presented transparently as templates for the user to implement.

agentic-os